Thank you very much to our witnesses who are here for our first panel of today’s evidence session. This is part of our inquiry into combatting new forms of extremism and we are very grateful to the three of you for coming to speak to us and taking our questions today. Perhaps it would be helpful for the public and those watching is if you could all each introduce yourself, and then we will go into questions. I will start with you, Milo.

I am Milo Comerford, I am at the Institute for Strategic Dialogue, which is an international counter-extremism think tank.

Hello, I am Adam Hadley. I am the Executive Director of Tech Against Terrorism, which is a UK-based NGO that works to save lives by disrupting the terrorist use of the internet.

Hi, my name is Imran Ahmed. I am the founder and CEO of the Center for Countering Digital Hate, based in Washington, DC. We have offices in London and Brussels, and we work on countering digital hate.

Thank you very much. We do not think we will be interrupted by a vote in this panel but, if we are, then please forgive us and we will suspend for 15 minutes at that point. I will start straight on the questioning and I will go to Chris Murray.

Thank you very much for coming in. I want to ask an opening question about how the extremism landscape has changed in recent years. Could you tell us a bit about how extremist behaviour has evolved and, in particular, how the role of social media and online tools has influenced that?

I am happy to kick off, if that is okay. At ISD, we have been monitoring online extremist behaviour and networks for over 15 years. We have been tracking regular transitions among the ideological threats from extremism, the different platforms and technologies that are utilised, and the different violent threats that are posed from both online and offline sources. We have been monitoring, in particular, far-right extremists and Islamist extremist threats for a long time. One trend that speaks to the terms of reference of this inquiry is a major growth that we have noticed in what we are calling nihilistic violent threats online. Those are a set of online harmful networks rooted in obsessions with violence that are not necessarily defined by a clear ideological framework or a set of political outcomes that you would traditionally associate with extremism threats and, indeed, the UK Government’s extremism definition is rooted in. These nihilistic violent threats might appear outwardly similar to extremism. They have similar impacts. They have similar online aesthetics. They often are actually quite proximate in online spaces to established extremism threats but they lack that ideological dimension that underpins extremist attacks. I think the case of Axel Rudakubana is very illustrative of this from last year, where we saw an individual engaged in online activity that reflected a volatile mix of nihilistic violence, of antisocial resentment, and sporadic engagement with broader violent online subcultures. I will give you two examples of these specific trends. The first is a movement that we have been monitoring called 764, which is part of the wider comm network. It started by engaging in sexploitation and the production of child sexual abuse material, but we have noticed a major pivot towards inspiring young people, and very often minors, to carry out acts of mass violence, which look very similar to terrorist attacks but are guided by a deep nihilism. Indeed, one of their subgroups is called No Lives Matter. It speaks to that nihilistic element of their ideology. The second is a rising threat from a group known as the true crime community, who are a set of online communities who obsess about mass killers and, in particular, those who have engaged in school shootings. We have unfortunately seen a major rise in cases linked to this community of violent attacks, including in the UK, but also especially in the US, which are inspired by the aesthetics of school shooters, like those that carried out the Columbine attacks, and are highly radicalised by communities of people where they are trying to impress them and trying to valorise them. They have hugely detrimental impacts, including a rising wave of violence in schools.

Before we open up the same question to the other panellists, you mentioned comm networks. Could you define comm networks for us and give us a little bit of explanation of what they are? We have heard it in the evidence and I would like to hear more about it.

Absolutely. The comm networks are online criminal networks at heart. As I mentioned, they are rooted in the production of child sexual abuse material, of sextortion, so often quite financially motivated in how they operate. They are a truly global transnational phenomenon. We have seen alerts raised by Five Eyes partners and EU partners around these networks that seek to exploit young people. However, the pivot that we have seen, and perhaps the most relevant thing for the terms of this inquiry, is this move towards nihilistic violence, which is a relatively new innovation within comm networks. We see people who are groomed into these communities, starting with essentially being ordered to carry out acts of vandalism and petty arson, and then swiftly moving up to carrying out stabbings, to sexually abusing and exploiting others, and indeed becoming victim perpetrators. There is this network of online groomers who are increasingly moving into extremism-adjacent spaces.

Thank you. I will expand to the other two.

Thanks. Stepping back a bit, I think the most important thing to stress is that the main threat that we are facing in this country comes from violent Islamist extremist organisations and individuals and the extreme far right. I would completely agree with Milo, of course, that the threat from nihilistic violence is significant. But it is really important, I would say, to focus on the existing threats and existing ways in which terrorists and those who have been radicalised are using technology in particular. The main things that have changed, I would say over the past, let’s say, five years—and I am sure Imran will agree on this—is that platforms seem to have largely given up in being proactive in finding and removing terrorist content. So the standards across the tech sector have dropped. There are a number of reasons for this. They have fired many of their teams, they have turned off tools. There is also algorithmic amplification. Platforms, of course, profit from sharing content that can elicit emotional reactions from people and therefore increase the chances of people clicking on adverts. In some cases, there are platforms that are actively promoting violent extremist content. The background here is that, unlike five years ago or even 10 years ago when Tech Against Terrorism was established, there is a sense in which some platforms are deliberately doing nothing. Other platforms are deliberately promoting this sort of material. The other major trend I would say is we are clearly living in a multipolar world where there are a number of significant geopolitical events. It is no longer the case that terrorist groups are well defined. As Milo was saying, what we are seeing is a mixture of different types of threats coming from nation states, coming from organised criminal networks, coming from transnational hate communities. Online what we are seeing is a mixture, not only of ideologies, but of grievances and actors. In some cases, we are seeing hostile nation states pretending to be terrorists, promoting extremist content—encouraging people to graffiti mosques in London, for example. From an analytical perspective, it is much more difficult to understand what is going on online and also our capabilities as NGOs, as Governments, to understand online content is pretty limited. The third trend, I would say, is generative AI. We hear a lot of hype about AI. It certainly has not done much to improve the productivity of our economy in the UK, but what it has done is improve the productivity of terrorist production of content. On its own this is not significant, except for the fact that if platforms are doing so little, if there are a number of threat actors all exploiting the internet, AI can be used to create a very large amount of content to overwhelm systems. Overall, I would say we are actually in a worse position than we were even five years ago, with many more threats. It can be very confusing to understand and analyse. I think Governments across the world are really struggling because there are no longer neat categories. Indeed, where the comm networks are concerned, it is not just terrorist content and child sexual abuse material, it is hacking, it is extortion, it is all these things mixed together—which I imagine, from a policy perspective, is very difficult to deal with.

If I can start right at the beginning, I think it is worth stating this for the record. The reason why this really matters, why online spaces really matter, is because the primary locus now of where we share information, where we set our social mores, our norms of attitude and behaviour, where we negotiate our values, and where we negotiate even the corpus of information that we call facts, are online spaces. Ten years ago when I started CCDH, sitting in Portcullis House, when I was talking to people and said that social media was resocialising the offline world, they told me I was ridiculous, because of course the offline world and the online world were completely separate. But we now know that that is precisely what is happening. I think increasingly people understand the role that platform decisions play, whether it is their algorithms or how they enforce their community standards—their purported community standards, the ones that they have in place to satisfy their true customers, their advertisers, that their content is not appearing next to radicalising or pornographic or other types of material. They simply do not enforce them. There have two real major changes that have happened in the past few years. The first one is the relaxing of standards, which was initiated by X and now has been followed by Google, YouTube, by Meta’s Facebook and Instagram. TikTok has not done anything yet, but I can imagine, given its new ownership, that it is likely to follow suit. The second is, as Adam correctly identified, generative AI. It is important to understand the economics of social media platforms. It is unprecedented in human history to have a mechanism by which you can send one additional message to one additional person for zero marginal cost. That produces an opportunity for distribution of content that is unprecedented in its scale. It is the nuclear age of disinformation and hate. The only cost to the producer, to the person trying to distribute hate material or the lies that inextricably are interlinked with hate—the conspiracy theories, the slurs, the libels, whether it is the blood libel or anything else—was that the cost of producing that content was high. Actually producing good quality viral content is a real skill. Generative AI reduces the cost of that to zero as well and creates a perfect feedback loop where you can produce content, test it on social media platforms for engagement, and then refine it based on that as well. The funny thing is that we are moving from one agent to another with generative AI. I do not think we have come close to comprehending what happens next without intervention. One of the issues that we have is that social media companies are turning a blind eye to rising levels of extremism, the cross-fertilisation of extremisms by their algorithms that leads to the hybridised, unclear ideologies that we call nihilistic now. The issue is that Ofcom is doing nothing to address it. We are two years from when they were given those powers, when the Act received Royal Assent, and four years since I was the first witness to give evidence for the Online Safety Act Draft Bill Committee. Twenty-one investigations into 69 websites: that is what Ofcom has done in two years, and not one into a major platform. Not one.

We will go on to specific questions on Ofcom, so we will cover that later. Chris, did you have anything else?

You have all mentioned several times about nihilistic extremism. Would I be correct in drawing the distinction between that and ideological extremism? Can you talk a bit about what narratives fit into nihilistic extremism? Some of the evidence we have heard is that common to everything are undercurrents of misogyny, of conspiracy theories, of antisemitism. That is across the piece, wherever you find it. Do you agree with that? What would your analysis be of what narratives drive this nihilism?

I will speak a bit about those three, which are very important, but also violence as an end in itself is, I would say, the unifying element across all of these. The difference with ideologically-motivated violent extremism is that violence, whether you agree with the cause or not—and by definition it was a supremacist, antihuman rights ideology—was to achieve a political and ideological end. The difference with nihilistic violence is that end does not exist. It is violence for its own sake and often for notoriety within online communities. It is a fundamentally misanthropic phenomenon. That is quite distinct from the types of extremism associated with the far right and Islamism, for example. But certainly those ideas of misogyny, antisemitism and conspiracy theories are at the core of what we are talking about here as well. Misogynistic ideologies, as all of you will know, permeate all forms of extremism. Misogyny is fundamental to far-right extremism, far-left Islamist extremism. It is at the heart of what those ideologies run on. But within nihilistic extremism, misogyny is often the very thin politics that actually galvanises violence. While we have not had a misogynistic act of terrorism in the UK, there are lots of acts of mass violence in growing numbers, which seemingly speak to a blend of very personalised grievances and an overarching misogynistic framework. Research from UCL has shown that misogyny in fact predicts violent extremist intentions, willingness to engage in interpersonal violence and support for violence against women. Misogyny is in many ways a gateway between all these different online communities we are talking about. When it comes to conspiracy theories, those again are fundamental to the construction of the world views that we are discussing today. ISD did work for Ofcom when it was first scoping its regulation looking at over 700 UK-based accounts. We found intense, fundamental hybridisation between extremist, hate-motivated and conspiratorial actors online. They occupy the same online spaces across platforms. That is a fundamental consideration for when we think about policy responses here. Conspiracy theorists themselves are increasingly linked with violence. We saw after the covid pandemic more and more cases of individuals who are radicalised as part of an attempt to overthrow Government through violence and so on. Ideology does not necessarily need to be present to motivate violence. Then, finally, antisemitism, which is in many ways the most fundamental phenomenon that is symbiotic with extremism across the ideological spectrum. Indeed, earlier in this month, after the terrible attack in Manchester on the synagogue, we found that antisemitic slurs on X increased by 28%, including over 6,000 posts accusing the attacker of being a false flag. Indeed, X’s AI assistant, Grok, provided speculative commentary supporting these false flag theories reinforcing antisemitic tropes. Indeed, since 7 October, we have seen that antisemitism is actually bringing together extremists across the ideological spectrum: Islamists, far-right extremists, far-left extremists making common cause around their shared antisemitism. In a way, a very useful phenomena to understand is some of the new ways that we are seeing online communities crystallising around this violent harm. Ideologies are still relevant. We still have a major threat from Islamist terrorism, as we saw with the Manchester synagogue attack. But more and more, it is about these shared narratives that are bringing together unlikely alliances within online spaces and are driving the violence that we are seeing both here in the UK and internationally.

I will ask one very quick follow-up. It is related to online spaces and back to comm networks. Is age a factor in this? Are we seeing it particularly prevalent among young people—I know we will come on to this—particularly on those comm networks?

I can hop in with some data points that we have at ISD on the predominance of younger people engaging in these spaces. These are exceptionally young profiles of people that are involved in this. We are seeing a number of susceptibilities among young people in some of the most recent data. The data is quite spotty, because at the moment, we only have data on those convicted under the Terrorism Act, so extreme right wing and Islamist offenders. We are seeing the average age of children convicted under this at the age of 15, and the youngest being 13. Among the risk factors we see in court records, we see a fundamental immaturity coming through in all the cases. Thrill-seeking behaviours are at the heart of many of the radicalisation trajectories we are identifying. Isolation, with a big spike in cases of isolation over covid, is at the heart of attempts by young people to showboat to peers through these expressive acts of violence. Adverse childhood experiences are often at the core of many of these youth radicalisation dynamics. There is also much more research needed on the link with mental health and neurodiversity. We are seeing again and again individuals with formal diagnoses of neurodiversities coming up in the caseload. We will get to it later, but I wonder if some of the ways that the systems for triage are based at the moment mean that people are being swept up in systems that were not built to deal with these challenges. Certainly, this is a large part of the youth radicalisation threat that is characterising the challenges we are seeing today.

That links very neatly into Margaret Mullane’s questions. We will definitely come to you, Mr Ahmed, on that.

Thank you for coming to the Committee today. We went to the National Crime Agency, where this issue was discussed. They gave their perspective and some case history to us. I have to say, personally, it was one of the most disturbing things I have seen for many years. I do not know if the public are as aware of this as they should be, which goes to Mr Ahmed’s concerns and how we deal with it. My question is more about the vulnerability. Milo touched on it a little bit. Are there indicators that can say where somebody is online and they are passively consuming something? Where do you identify where they are likely to then participate and they make that move? Is there anything that you have identified, all of you?

Mr Ahmed, let’s start with you, because you were itching to come in, and then I will ask Mr Hadley.

I will make a point on young people. Mobilisation is the classic CV and CT space and I do not want to comment on that, because it is not what we do as an organisation. We look specifically at digital spaces. We did a survey of 14 to 17-year-olds in the US and UK, looking at both children and adults. We asked a bank of conspiracy questions. To what extent do you think there is any truth to one of these conspiracy theories? Across both countries, the US and the UK, the most conspiracist cohort, and in fact, for every single one of the nine conspiracy theories—a trans conspiracy theory, climate conspiracy theory, deep state conspiracy theory, antisemitic conspiracy theory, misogynistic conspiracy theory— across every single one, 14 to 17-year-olds are the most likely to believe that they are true, which is unbelievable. It is a sudden inversion in what we expect to be true. We expect it to be your crazy uncle, the one who sends you the WhatsApp message with the crazy conspiracy theory. It is the 14 to 17-year-olds now. I think that in itself is extremely disturbing. The other thing that we are seeing is that, while we call these ideologies nihilistic, and I do not have an issue with the use of the term, I think it is really important to understand they are hybridised extremisms comprised of a series of complex and incoherent beliefs that could not credibly be called an ideology and are replete with internal tensions. But then, so are most people’s opinions about the world around them, and they somehow manage to rationalise them. What underpins all conspiracy theories, however, is a belief that you cannot trust the sources of information that you are currently getting. In our very well-received report, “The Anti-Vaxx Playbook and the Pandemic”, we went through millions of messages about vaccine disinformation. There are only three messages: “Covid isn’t dangerous. Vaccines are dangerous. You can’t trust doctors and the Government.” Every conspiracy has that third element to it, which is about epistemic replacement. The funny thing is, they do like the Government when they control it, but these are authoritarian, often extremist forces that are seeking to undermine the status quo. I will hand over to Milo.

I will come in on the risk factors.

Before you start, you talked about 14 to 17-year-olds, do we think that the lockdown and the age of those young people at the time of lockdown has had a particular impact on them that we are now seeing it in 14 to 17-year-olds because of the age they were when we went into lockdown?

It is impossible to know for sure, but most experts in the area would suggest that the isolation created by covid must have contributed to this. But there has to be something else at play here. To the point about vulnerability factors—I am sure Milo will add a lot from his perspective at ISD—there is no one radicalisation or mobilisation journey, which is what makes this so difficult to predict. That is what makes the job of counter-terrorism police so difficult. But there are commonalities, there are common vulnerability or susceptibility factors across the board, whether we are talking about violent Islamists, extreme far right or those who belong to the nihilistic extremist communities. Those sorts of things are often focused on safeguarding, individual vulnerability, their background, potentially neurodivergence and, of course, grievance narratives. All that can be mixed in various different ways. Depending on the specific circumstances of the individual, it may result in them becoming a particular type of extremist or terrorist. The public health or public safety model is clearly the way to go, trying to understand what are those vulnerabilities that an individual may have and then working from there. In a sense, focusing on ideologies can be quite misleading, because it is never really the case that ideology alone will radicalise or mobilise people. There are always circumstances and, crucially, there are always signals and indicators of online activity. With Jihad Al-Shamie, for example, I think there is an awful lot in open source about his radicalisation journey, with dozens of potential red flags, many of which could apply to different types of terrorists as well. Focusing on the detail of how people behave online is really important, but actually the dataset about this is really limited. There is relatively little data publicly available about how convicted terrorists use the internet. At Tech Against Terrorism, that is something we would like to see more of, because we know that in every, or almost every, case of terrorism in this country over the past five years, the internet has played a fundamental role—not an incidental role; it has been fundamental. Yet our understanding of that journey, what sites people have been on and what content they have consumed, is extremely poor. If there could be more research into that, I think that would be really useful for practitioners in this space. Also, it could be used to advise tech platforms on how to detect these signals before people commit violence.

Equally, you would also, if you wanted that research—and I agree with you, it needs a massive piece—you would need to understand those who have consumed that and did not go on that journey. Again, that is a massive piece going forward that is needed.

Mr Ahmed, can I push you on something you said then? Is it not utterly predictable that 14 to 17-year-olds would be the most vulnerable group to conspiracy theories? They are young and impressionable, those who will believe simplistic notions they see on the internet. If that is the case, is the Australian example of limiting access to that age group the way forward?

I will let you think about that and we will bring Milo in.

I completely share that concern from your NCA visit around the terrifying nature of this and the fact that it is so misunderstood by caregivers, by practitioners on the front line. This is again and again coming on our doorsteps; people need practical resources for the kinds of susceptibilities that you are talking about. The backdrop to this is that the radicalisation trajectories of individuals in these spaces are so much faster than that of people who joined ISIS, for example, those who joined far-right extremist groups. We are talking about days and weeks, rather than months and years as was the case for the previous generation of radicalisation. We have produced a resource that we will be releasing soon and will share with members of the Committee. The release looks to speak to the specific cocktail, if you like, of factors that are bringing people from passive consumption within these spaces into real-world mobilisation, having seen more and more cases of this in our work. We have distinguished between indicators, which are signs that someone might be part of a subculture of nihilistic violence, which might be things such as extreme interest in gore websites, an expression of a desire for notoriety or wanting to be well known in a community, or an obsession with mass killers. That in itself was necessary but not sufficient for taking action. We see a number of accelerants, which are co-morbidities, if you like, with this subculture involvement, which moves someone more towards violence including self-harm and examples of suicidal ideation; eating disorders very often come up again and again in combo with these subcultures, as well as psychosis and isolation, which we spoke about earlier. The final thing I will add into the mix is triggers. Very often there is a destabilising event that happens which increases the risk of imminent harm by people to others, but also to themselves. Here we saw again and again sudden life changes, social stresses, financial crises that were presaging violence. In that case there is this intricate layering of things that happens within these online communities but, as Adam said earlier, it also produces a range of different opportunities for intervention, for recognising risk and for engaging in online spaces in a much more sophisticated way to reach those who are vulnerable, which hopefully we can speak about a bit more later.

On that point, I do worry about having such a broad array of potentially mobilising factors. We studied incels in great depth, so let’s say you are an incel and you believe that women are basically malignant, evil, and less than human. It could be someone not giving you the right change at the supermarket. It could be someone giving you a funny look. It could be someone that you went up to and said, “All right, darling” and they said, “Get away from me”. It could be anything. It could be your sister saying something mean to you. It could be anything. What we do know is that these people have, over a long period of time, been able to consume, without interference—and in fact often accelerated by algorithms normalising the content that they are seeing—a vast array of hate-filled content and the conspiracy theories and lies that underpin hate. That brings me to Mr Kohler’s point, and yes, you are right. T2A is an idea from criminal justice, but in the transition to adulthood between 14 and 24, until 14 you are essentially being socialised by your parents. After 24 or so, you are being socialised by partners. Between 14 and 24 is that “Lord of the Flies” moment when you are being socialised by each other. Those intense pressures as you are trying to identify who you are as a human being are sociological. There are neurological changes. It is an era of great frontal cortex plasticity. There are a lot of changes happening in that person and then you inject into that completely destabilising ideologies, content, stuff that makes them feel bad about their body. CCDH does not just look at hate content. After six months, we were looking at antivaxx content and since then we do a lot of work on children’s mental health, eating disorders and self-harm content. They are being exposed to all of that, which causes this incredibly destabilising effect on young people and their identity, their sense of self, their sense of where they fit in. Then you are throwing in content that is blaming particular groups for that. You can understand why we are in a crisis right now. Australia has pulled the emergency handle, right? This is a factory floor where the products are killing kids and killing other people, and they have said, “Okay, push the button. Stop. No more until we work out what the hell is going on”. We have chosen another approach, which is we passed the Online Safety Act. It theoretically gives the powers for Ofcom to enforce rules that platforms have to be safe for children, that they have to think about and show that they have done adequate risk assessments, they have red-teamed their platforms. Unfortunately, Ofcom has not really got going yet. I would argue it has failed catastrophically in its first two years to get going on these issues. So I can understand why you might be thinking that maybe we should do what Australia is doing and push the stop button.

I am very keen to get an understanding from all of you of how effective you think the Prevent model is in identifying young people as being radicalised online specifically. Would you agree with some of the written evidence that we got as a Committee calling the online space a blind spot for Prevent, noting that there is no referral point for individuals who are displaying problematic behaviours online?

I would agree with that generally, but I do not think it is the fault of the police necessarily. I think their resources are somewhat limited and there is this overall perception, as I was saying earlier, that the internet somehow is incidental to terrorist attacks, whereas it is really fundamental. I would be curious to understand what the resources are across Government to proactively search for content online. The Online Safety Act and Ofcom’s remit is pretty passive by design. It is not their job to go and search for content, but it does not appear to be anyone else’s either. It seems to be left to individual citizens to report content or to NGOs to do so. The platforms have prioritised the profit and socialised the negative impact. They have pushed responsibility to proactive searching on everyone else. I would certainly say that Prevent and the police needs significantly more resources. It is not just in this area, though. It is also detecting child sexual abuse material or online scams. There is not one single approach across Government to find harmful content online, which is quite absurd considering there are so many crimes being perpetrated online. There is no single entity or budget to look at content. At the same time, platforms have made it significantly more difficult and expensive to find material. So it really is a blind spot, not just for the Government, but for researchers. Something does need to change, because I cannot see how we can protect our citizens if we do not know what is on the internet and we do not know how to hold platforms to account. As it stands, we have to trust platforms, and I am not sure about you, but I do not trust them. I do not trust that they have our democratic interests at heart. I trust that their priorities are their shareholders and, quite evidently, they are profiting from polarisation. The question I would have for the Government is: what investment are you making to understand how the internet is being used by adversaries? I would also say it is not just extremists. As I mentioned earlier, there is a significant amount of evidence that traditional hostile nation states are also sharing material online. At Tech Against Terrorism, we found examples of fake far-right groups apparently trying to pay people to graffiti mosques in London and across the country. This activity is all over the internet and we are really worried that there is not any systematic understanding of that.

I will come in on the Prevent programme specifically. Certainly, I agree that online spaces are a major blind spot. You might have heard an anecdote from Lord Anderson when he was at this Committee that in the case of Axel Rudakubana, one of the referrals was on the basis of a teacher—referrals to Prevent, that is—looking over his shoulder at online activity that he was engaging in in the classroom—an extraordinary, incidental opportunity to see what his online activity was—and saw this harmful engagement with genocidal content and so on. Without a doubt, we need to be embedding and mainstreaming considerations of the activity online within how we assess the risk to individuals. More generally, on Prevent’s ability to respond to these threats, these hybridised threats that we have been talking about have ended up on Prevent’s plate not by design, but due to the absence of other support systems that exist to deal with this specific challenge. There is an inappropriate triage that is happening. In last year’s statistics from Prevent, the largest category of cases was vulnerability present but no counter-terrorism risk. That is fundamentally a category error. What that suggests is that there is a miscellaneous category that is larger than any kind of coherent category. Only 6% of those were then referred onto Channel, which is the provision for interventions. We need—and I am hoping we can speak a bit more about it—a fundamental recalibration to mitigate extremist-related violence, which moves beyond just ideology, as we have heard earlier, towards comprehensive violence prevention and a framework that is rooted in a public health model, which I think is, in many countries, showing to be an effective way of addressing these issues.

We will be coming on to that in a moment.

What do you feel could be undertaken for the Prevent model to be improved, specifically to capture new forms of extremism and identity of those who are undertaking some of the things that you have been talking about in an online space? Maybe while you are thinking about that, what are your thoughts of the recommendations that came forward from Lord Anderson, the 10 recommendations, and do they go far enough?

I thought the recommendations were absolutely on point from Lord Anderson. We were very pleased at ISD to be able to help contribute to those as well. There were two main ones that stood out to me. The first were around the widening of Prevent referrals to include this category of violence-fixated individuals. In the short term, that would go a long way to capturing those who are representing real risks like we saw in Southport, but also in a number of cases up and down the country. The second—and this is where we come in with the public health model—was this recommendation around creating a much more holistic approach to violence prevention, which joins up efforts that are happening across Government in a much more unified way. When I say a public health model, this is a non-securitised, ideology-agnostic approach to violence prevention, rooted in disease prevention principles and boosting protective factors. It addresses risk factors and encompasses a spectrum of interventions from primary prevention upstream, working to inoculate and build resilience, right through to downstream efforts that are about reaching those who are vulnerable, off-ramping them and providing them the resources they need. What is very effective about this is that it can knit together these efforts that are happening across Government and bring together public safety, health, social care, education, into a much more unified approach. This has been quite successful in some cases.

We will come on to this shortly, we have some specific questions on it. I am glad you have faith in intra-Government working. To the other panellists.

If I may make one suggestion, it is this. My recommendation would be to create one single entity that centralises the capability and capacity to search for this content online—not just terrorist content, but child sexual abuse material and online scams. At the moment Prevent officers, who are well-trained police officers, are expected to be able to search the internet for content. I think that is unrealistic. There seems to be no money to invest in the technology and data access or the training. Therefore, I think centralising and not just focusing on one harm, but across various harms, would be really helpful.

To make the point, I served on the Commission for Countering Extremism under Sara Khan and alongside Sir Mark Rowley on the Pilot Task Force Steering Committee. It was doing some quite interesting work on the upstream digital drivers of extremism and trying to identify them, trying to go through them in detail. I thought it was the only time I interacted with an organ of government where I thought, “God, they actually get it; they really, really get it”. It is an absolute tragedy that it has been mothballed and that its institutional structure and expertise are going unutilised right now.

Chris Murray and Peter Prinsley have some quick follow-ups and then we will move on, I promise, to the public health model.

This is a question we asked our previous panel. I understand Prevent has its flaws: there is a lot of mistrust around Prevent in the community and the extremism picture is totally changing, as you have set out. But one of the advantages of it is it is quite embedded. People know what it is. Something that schools and social work and the police can engage with. What is your analysis of what I have just said? Is that really a value of Prevent? I am worried that if you totally reform it you will throw the baby out with the bathwater. Is that a legitimate concern to have or should I be more relaxed?

Before you answer, Peter. Q60            Peter Prinsley: My question is about AI. I have been sitting listening to this with growing horror, frankly. It seems to me that we are at a moment where we might have computers that can deal with some of this stuff. You were talking about AI generating some of the content. How can we use AI as a filter? How can we persuade the tech companies to filter some of this rubbish that is being generated and use AI itself to our great advantage?

I am giving you some thinking time, because we have got questions on AI as well coming up. To give you some thinking on that, but perhaps if you can cover the preventive things.

Perhaps I can give a very, very quick reaction. You cannot persuade them to do it, because the controversial content, the hate, the extremism are drivers of revenue. They create contention, which creates attention, which keeps people on the platform arguing with each other. It is highly addictive and unique content to their platforms and they absolutely do not want to get rid of it. That is why they have done quite the opposite: they have reduced the volume and they have said, “You know what, you should put a community note on it. That is the best way of dealing with extremist content.”

The technology is so sophisticated these days that, if the platforms wanted to detect this, they would be able to do so, because their entire business model is predicated on developing comprehensive psychological profiles of all of us in order to sell things. If they can figure that out, they can absolutely filter material and nudge people towards—

When you go home tonight, I want you to record a video of yourself and put five seconds of Taylor Swift’s latest song in the background and try uploading it to YouTube and see how fast their algorithms can identify her song and shove it straight down. They have found ways to look at it. They are looking into the audio file, they are looking into the content, they are matching it against hashes, they understand exactly how to do this. They have been doing it for a decade when it comes to copyrighted material, and yet when it comes to content that might make our kids cut themselves or might lead to a terrorist attack, they somehow cannot do it.

We will come on to the AI and the tech things, but can we have an answer to the Prevent issue that Chris Murray raised? Then I will come to Lewis Atkinson on public health.

I would say on Prevent it has been remarkably successful. There are not that many terrorist attacks, and the ones that do happen are unusual in various ways. I would broadly say that Prevent is a successful model, though somewhat under-resourced.

I will add to that. I completely agree with not throwing the baby out with the bathwater, but it is about contextualising Prevent within a broader set of safeguarding apparatuses that speak to each other on the front line. I think that is really important. We have made suggestions around the ways that you can integrate some of the existing multi-agency safeguarding hubs that exist at a local government level with Channel boards to be able to create integrated risk assessments, the Prevent oversight boards that exist with serious violence boards and community safety partnerships. These are all existing infrastructures that currently exist in completely siloed spaces locally, which mean that individuals are slipping through the gaps. Prevent purports to be a safeguarding mechanism. It needs to be much better at safeguarding against violence in an ideology agnostic way in order to prevent things like the Southport attack from happening.

I apologise that I keep interrupting, but I am conscious of time and I do want to get through all our questions. Lewis Atkinson will now ask about public health.

Thank you. Mr Comerford, you have mentioned this already, and I know you covered it in your written submission and ISD’s written submission as well, but could you tell us more about what you think a public health approach to tackling extremist violence would be in practice, and particularly what you think Government should be doing practically if they want to adopt that?

Yes. Practically speaking, I talked through the principles a minute ago and why I think it is a very effective model for thinking about these issues, and why the analogy, if you like, with broader public health programming is very effective when you are thinking about addressing violence. But, practically speaking, this is about a whole-of-Government approach; I am sniggering because of the earlier remark that throwing that out there and saying it needs to happen is not particularly effective. What I think will actually be more impactful in embedding public health approaches to violence prevention is getting to a local government level. I think it is really important to be able to set direction from national Government, to be able to have the relevant Ministries that I outlined earlier speaking to one another and taking ownership of elements of this, but the proof will be in the pudding if this is embedded at a local level. We have an ability to do this at a city level. The Mayor of London has his own counter-extremism framework that is about rooting attempts to tackle extremism in the specific risk factors in London. That is about understanding what the emerging risks are, analysing the threats online and offline, and creating a series of programmes to be able to respond to that. The same holds in Manchester. I think what will be really important is getting, as I said earlier, with some of the multi-agency safeguarding hubs, the Prevent oversight boards that exist and deal with very small parts of this problem to be able to integrate into a much more unified approach to risk assessment and then to triaging vulnerable individuals. I think that is at the heart of this. The direction needs to be set from central Government, but implementation really needs to get to the local level.

Are there international examples from countries that have tried embarking on this and you think have been successful?

Yes: on the local level, the Netherlands has a fantastic model called the Safety House Model, which has actually been running for many years now as a sorting house for these complex cases. It is administered by municipalities, and I think that works very well in a context like the Netherlands, which is very decentralised, but they are able to triage complex social referrals for individuals across a spectrum of different harms. That worked very well in the context of ISIS returnees and being able to assess risk and respond there. I have also just returned from Canada, which has a very effective model, Public Safety Canada. The Canada Centre essentially drives deep cross-agency co-ordination that goes beyond traditional counter-terrorism. Normally this is a very securitised function, but they bring together public safety, health and social care, education and broader considerations around online harms. In the context of this conversation around new forms of extremism, it is a very resilient system that is able to quickly adopt these threats into how they are assessing risks and responding in a way that, frankly, I think makes the UK look quite cumbersome in comparison. They are looking at these emerging challenges and they are able to pivot responses because of that cross-Governmental collaboration much more effectively than I have seen in the UK.

Do any of the panellists have any particularly different views about limitations of this approach or anything? Chair: I have Jo White and Paul Kohler who want to ask some follow-ups and then we will move on to the actual technologies.

I will bring Paul in with his follow-up as well.

You referenced London and Manchester, which I think is part of your Strong Cities Network. I can see what it is doing. What is the evidence that it is working? As a London MP, I am not aware of this.

I can take both of those quickly and then pass them to the other panellists. I absolutely agree with the framing around protection of young people, in particular. I think there is a real risk, when it comes to this correlation with neurodivergence and violent ideation, of essentially securitising young people with neurodivergences wholesale, which is incredibly damaging and actually a poor reflection of where the risks are. It is about these different risk factors being present together and often being triggered by specific—often non-identifiable, as Imran mentioned—real-life circumstances. On understanding potential risk, I think it is about recognising some of these features and indicators that we have been talking about today, so that you feel really confident in being able to navigate these challenges early on, rather than when it gets too late and you are dealing with a much more challenging situation of diverting someone from an act of violence. It is about building the evidence base. I know the Home Office is doing much more research on the relationship between neurodivergence and violence, so I think it will be really important to support analysis that can address that in a nuanced way that is strengths-based and looking to protect people rather than just securitise new cohorts. But I think that caregivers and those who are on the frontline understanding the specific risk factors that they need to be on the lookout for—it is very hard and quite challenging to isolate those from some wider behaviours that we talked about earlier—is absolutely essential to how we will solve for this. On the question around local responses. Yes, as you mentioned, our Strong Cities Network at ISD looks to embed good practice from cities around the world, in fact, around local responses to hate, violence and extremism. We are growing our evaluation work of this programme. We have now, for several years in a row, evaluated the Shared Endeavour Fund, the London scheme, which has grown and grown in its footprint across all boroughs in London on its support for local projects. One of the things that has grown there is moving from funding well-intentioned local projects that are about bringing people together and creating a nice photo op to monitoring the attitudinal and behavioural changes of people, in particular young people, engaging in this kind of programming. We have developed a standardised set of measures that allows you to look at the impact of a workshop, of an event, of the number of different interventions that are happening, and be able to standardise across all of the dozens and dozens of projects supported by the Mayor’s Office and look holistically at the shifts that have happened to in young people’s resilience to extremism and their understanding of political violence being an unacceptable outcome. I would be very happy to share those with the Committee; there are some promising evaluations emerging from that programme.

We have less than 10 minutes and we will need to cover the new technologies. This is incredibly interesting and we are getting an awful lot from it, but I am conscious of time and we have got another panel waiting so can we keep it short, please?

Thank you very much to all of you. I want to go back to something you were saying, Mr Ahmed, the way in which you described what people are doing with copyright. I have not heard it articulated like that before, so thank you very much; it was very interesting. This is not the first time you have probably said something like that. I wanted to understand from the perspective of all of you, since this technology has been in place for 10 years and we understand that those that host these platforms are making profit from it, what is the unwillingness from Governments to put policy forward that would actually ensure that they use the technology to prevent the proliferation of these really awful things online?

Having worked with the previous Conservative Government and all parties on the Online Safety Act, I would say this Parliament has done its job. It did it really well, in a process that also took the politics out of it by having a draft Bill Committee. I think that Parliament should be very proud of what it did because it was the first Parliament in the world to get to grips with this. Because of Brexit we were able to legislate on our own before the Digital Services Act even. It took a bit longer than the Digital Services Act to promulgate, but nevertheless we should be very proud. The issue comes now on what has happened since it was promulgated and the failure of the new regulator, Ofcom, to get to grips with the problem. It has done just 21 investigations into 69 websites, not a single one of which is a major platform for anything. Even with the evidence base we have been providing, for example, of the proliferation of antisemitism on X in the wake of the Manchester Synagogue attack, five miles from where I grew up, there is nothing from Ofcom—absolute radio silence. It is unacceptable.

I have two points. I am not sure members of the public quite realise how much obvious terrorist content there is. By obvious, I mean material produced by ISIS, with their logo, with their name. It is really easy to find this material. You could find an ISIS network on Instagram in 30 seconds if you knew what keyword to use. The tech sector will often say, “Well, all this content is very hard to find, it is ambiguous, and we must have freedom of expression”. Well, actually, there is a huge amount of really obvious material produced by designated terrorist organisations. To Imran’s point, it would be so easy for them to find this if they wanted to, but they do not want to. In defence of the Government, in one sense, the content is not illegal in most cases. Terrorist content is not necessarily illegal. There has to be intent. I would argue there is merit in exploring defining particular types of content as illegal—for example a terrorist manifesto or material produced by a terrorist organisation. I would argue that should just be illegal, but at the moment it is not. Ofcom’s role is not to adjudicate on individual pieces of content. That is not how the Online Safety Act was created. But perhaps there is an opportunity to develop legislation that does clarify what content is illegal, which could then be put to platforms as a legal obligation on them.

To reinforce that, I think it is really important to emphasise that platforms want to be talking about content. They are very happy talking about the amount of terrorist material they have taken down, how great their automated techniques and their AI tools are for removing this content. What they do not want to talk about are the fundamental systems that underpin the way this is served to young people in particular. We have produced research for over five years of this phenomenon of algorithmic amplification to young people. This is not just about content being accessible online to those looking at it; it is about it being served up to young people. You can set up an account as a minor and by engaging with pro-Hitler content on X, for example, you can quickly have the first 10 posts you see on your “For You” discovery page be pro-Hitler content without even following these accounts. It is really important, in the way that the Online Safety Act looks to do, to have a conversation around the systems that are promoting this content, rather than just talking about how we will take down all this material. We will not be able to remove the whole of the nihilistic new forms of extremism that we are talking about today from the internet. We have to truly understand the roles that different social media platforms are playing in servicing young people and creating communities around this kind of content. I think that will be a really important part of the solution.

To build on that, there is no mechanism in the Online Safety Act for taking down individual bits of content. It is an excellent piece of legislation that deals with systems and risk factors, and creates incentives for platforms to do the right thing—to do what they claim they are doing—and then has sanctions if they fail to do so. The problem is that those sanctions have to be credible, which means they need to be ready to use them.

There is a fundamental flaw in that, though: how can you evaluate the systematic ability of a platform to remove content if you are not monitoring and looking for how much content is out there?

Without meaningful data access as well.

The data access problem—we were the organisation that was sued by Elon Musk for the act of doing research on his platform. He said the terms of service of his platform say you can only use the platform to read it and consume ads, not to research it, which is bananas. So he sued us and we beat him in court. We have now got these statutory data access pathways through both the OSA and the DSA. The problem is that they are still not in force yet. Again, we just need people to get on with it so we can tell you what is going on and help Ofcom to help itself.

If I have time, I will come back to that, but I have to get through some other questions before I get in trouble.

Can we bring Jo White in very quickly then?

Well, I think we should be deeply suspicious of various narratives that we are seeing in this country. We need concerted effort to understand where particular narratives are coming from, especially anti-Muslim sentiment and anti-trans sentiment. These things do not come from nowhere. Often the indication is that some of this comes from abroad, but also there are lots of domestic sources for this. Fundamentally, no one really knows, because data access is so difficult, because there is no funding, because Government is not really looking. Maybe it does not want to know. The problem is that no one knows. No one is really looking. Platforms most definitely do not want you to be researching their systems. In the case of X, you are forbidden from doing so. There is a huge intelligence gap, especially considering the environment we are living in at the moment, where there is concern about war with nation states. We know that many of those nation states value information operations before everything else. We know that in hybrid warfare, information available on the internet is the first thing that they will seek to manipulate. Yet, in the Strategic Defence Review, the reference to the internet was almost minimal. It is almost as though, from a national security perspective, we think the internet exists on its own. We do not realise that the information environment is fundamental to not only hybrid threats from extremists, but hybrid threats from nation states. That is a real problem.

Can I check the panellists are happy to stay on for a few more minutes? Okay, Bell.

To pick up on that point about things going slightly beyond the internet. How has technology and the proliferation of extremist content online altered the radicalisation process?

Maybe you could also address the point that Peter Prinsley raised earlier around AI and the role it is playing and whether there is a role for AI to actually solve the problem—I realise “solve” is a bit of an extreme thing to say.

The fundamental shift is the one I talked about at the very beginning, which is ultimately that the radicalisation trajectories of these people that we are talking about today are geared towards online communities of harm rather than specific organisations or groups. When you previously were talking about a far-right group that you had a membership card to, or pledging allegiance to a group such as ISIS, with these communities you are talking about incredibly volatile and divergent online communities, where it is not really about belonging, but about participating in an environment where these things are almost more real to people than their offline worlds. That is a fundamental step change in the radicalisation to violence that we are seeing, which is not that someone is being groomed into an ideology, but rather they are participating in an online environment where violence is totally normalised and you get to a point where, unfortunately, a subset of those people engaging in that will carry out acts of real-world violence. Finding those individuals is very challenging, and I do not underestimate the task of law enforcement in finding the needles in the haystack. But, as we said earlier, it is fundamentally about the construction of platforms that are hosting these spaces, where this socialisation to violence is able to happen, and the pathways of radicalisation are fundamentally entwined with the systems that underpin social media platforms.

After 9/11 there was an approach used in studying how people become radicals, and it was the funnel. Basically, you start with being a Muslim, then you become an extreme Muslim, then you become a jihadist, and then you become a terrorist. There was a funnel there and it starts with the Muslims getting poured in and eventually you get a few terrorists at the bottom. That is no longer used, but that is exactly what we have on social media. We have a system in which people are able to be identified for psychographics—not necessarily their religion or their race or ethnicity or anything else; it is about the psychographics. Young men who have problems with women, for example, who have problems with their mental health, who have problems with suicidal ideation, who have problems because they are NEETs: I am using a specific example here because that is exactly what we found in our study, “The Incelosphere”. We found that for the world’s biggest incel site, the people who had set it up had set up three feeder sites. One site was a NEETs forum, one was a suicide forum for people with suicidal ideation and mental health problems, and the other one was a looksmaxxing forum, which is for men with extreme body dysmorphia. These people had set up these forums as feeders to identify young men. The looksmaxxing forum—looksmaxxing is a term that is used by young people on social media—was the number two or number three site on the Google search. If you searched for looksmaxxing, you would get to this forum; it had very good SEO. Once you were there, you were being identified for potential recruitment for the incels forum. In the incels forum, which at the time when we were studying it was 16,000 men, was the space in which you were inculcated into the full set of incel beliefs, including that raping children was an acceptable means of ascending from inceldom, which is a term they use. You are no longer an involuntary celibate if you have sex, so they had agreed, in that year that we were studying them, that raping children was an acceptable form of ascending from inceldom. That is how that funnel works. The funny thing is that that old-fashioned, discredited, slightly embarrassing funnel approach that was being used 20 years ago is exactly how it works on social media.

I will bring this to a close at 3.15 pm because I am conscious that we have panellists waiting.

Just a question about algorithms on that point, the role that they are playing on pushing extreme content to users, and how much control the Government have over that—and AI; we might go back to the AI. If one person can answer the algorithm question and one person can answer the AI.

Let me bring Paul Kohler in as well because he has questions on Ofcom and we can do them all.

I want to ask a very basic question. Is it the regulator or the regulations that need changing? Is it Ofcom, is it the legislation behind it or is it both?

The regulator is failing chronically right now. The regulation itself was a very, very good first start. We do need iterative reform over time and we need to see if there are any problems that it causes, but we also need to deal with AI. We need to deal with algorithms and transparency of algorithms. We need to deal with advertising and transparency of the advertising tech stack, understanding the degree to which economics influence the distortion of the lens that social media provides on the world. Those are things that we could add to the current legislation, but we need to test how the legislation that we have in place works in practice. You cannot test how it works in practice, though, if no one is doing anything with it. Until we see it in practice, we will not be able to tell.

Why is Ofcom not doing anything? Why is it not using the powers that it has?

I think that Ofcom, correctly as an independent institution, is trying to set its own pace and trying to build up its awareness. But it has spent an enormous amount of money, recruited an enormous number of people, made an enormous number of promises, and is failing to engage with civil society. I can tell you that the only time I get a phone call from the CEO of Ofcom is after I have criticised her on “Sunday with Laura Kuenssberg”. Apart from that, they never bother. These are the problems that we have.

You are expecting a call later.

I am—I actually saw her deputy when I was in Montreal last week. We want to see some action being taken by it, especially looking into the major platforms. The other thing is that the major platforms have been getting worse in the last year. They have all been reducing their safeguards, and they are doing that because they think that there is no one watching. They think that there is no credible threat of sanctions. The whole point is always to be investigating, because there is always something to be looking at. If you do not, you are doing the opposite of placing pressure on them. You are telling them that they are getting off scot-free.

Peter Prinsley has a question, I know, so I will bring him in, and then if we can wrap up quickly.

I will ask about research, but somebody said that we are looking for needles in the haystack. I think that we have to keep looking at the haystack. It is a haystack. Are there are ways of scrutinising what individuals are seeing? At the end of the day, this is individuals looking at screens in bedrooms. I simply do not believe that there are no technical fixes that allow states, or state authorities such as police services, to look at what people are seeing and somehow use this to filter out who are the people who are likely to be at risk. It seems to me that we must be able to do this.

If one person can take everything that is left over and deal with it in 30 seconds, that would be marvellous.

To come back to the haystack point, which I think is absolutely right, looking at this purely through the lens of how to reach individuals at risk is slightly missing the picture of what we are seeing here. What we are dealing with is a mainstreaming of these phenomena. That requires understanding how entire online communities are shifting towards violence, misogyny, conspiracy theories and so on, and thinking about what approaches will be effective in altering that trajectory, whether they are regulatory or new interventions that are able to get to this at scale. We need to be thinking about piloting interventions that can shift the gravity of communities online, because you can spend your whole time looking for that needle in a haystack but, as you say, the haystack itself is shifting. We have heard already about that in the context of platforms rowing back on their commitments and doing less and less. There is a fundamental shifting of the ground that requires, exactly as you said, looking at the big picture.

Quickly, if I may, to the filtering point, schools and universities have an obligation to ensure that their students are safe online. There is a question as to whether they are using the right technology and are evaluating it correctly. Yes, it is possible for schools, and they are clearly not doing enough of it. On the business model point, it is less about AI and algorithms and more that the fundamental purpose of social media is to generate advertising revenue. So the system itself creates emergent threats. Not only that, but you have content creators who are incentivised to create content, to create sites and so on to make money from advertising. The system itself is generating these risks and these threats in many cases and it is about the business model.

If I can make a point, on average, 14 to 17-year-olds spend four and a half hours a day on social media. Every hour or two, the equivalent of the entire Netflix library is added to YouTube. You do not understand the scale of what we are talking about here. These companies are enormous. It is not just the inputs into it, but also the outputs. Take every single 14 to 17-year-old and multiply by four and a half hours. There is no human-based system that could possibly analyse it. We do studies internally at CCDH where we look at what happens to an individual. We model an individual going down a radicalisation pathway to show how it works. That is exactly how platforms do the studies themselves internally, we have discovered. The truth is that we have this relatively impossible, chaotic black box that no one really understands. We have unleashed it on the world with zero oversight, zero accountability and zero transparency and we are wondering why things have gone wrong.

I will have to wrap it up, although we could probably spend all day with you. If you have anything further, we would welcome more written evidence. Perhaps if, as a Committee, we think of more questions for you, you would not mind perhaps if we can write to you. But I will have to call it to a close, I am afraid, because we do have a second panel, and I believe that we will have a vote at some point soon as well, which will get in the way. I will call to close now. Thank you again.   Witnesses: Laurence Taylor and Paul Giannasi OBE.

We now move on to our second panel. Thank you so much for being here. Can I ask each panellist to introduce themselves?

I am Laurence Taylor; I am Assistant Commissioner in the Metropolitan Police. I am the head of counter-terrorism policing, which is a national network, a collaborated network, that exists to protect the public from terrorism and hostile state threats. Across England and Wales, it is a network. We also have strategic relationships with the Police Service of Scotland and the Police Service of Northern Ireland. I lead our collective response to that across the four Ps of the CONTEST strategy—Prevent, Pursue, Protect and Prepare. We will talk about Prevent, I am sure, in due course.

I am Paul Giannasi. I am a police officer by career. I spent 12 years running the Government hate crime programme, first as a seconded officer and then as a civil servant. I now advise the police on hate crime, working with the national policing lead for hate crime. Within that, I run some of the functions that exist to support national operations. I run, for instance, True Vision, the online web facility, the Online Hate Crime Hub, which takes the first response to some allegations of online hate crime, and I also chair the National Community Tensions goal group, which assesses the levels of tensions in society, assesses the risk and agrees actions to try to mitigate those risks.

Thank you very much. We are looking first at the police approach, or law enforcement approach, to tackling extremism. We will start with Paul Koehler.

Your remit runs to hate crime, extremism and counter-terrorism. Can you show us how you navigate the threshold between those three?

There is a risk with all of this that we conflate a number of things. The legislation on terrorism is really clear. It is terrorism if it is there to influence a government and it is based on an ideology. Extremism might be part of terrorism, but not all extremism is terrorism. You have this huge spectrum where extremism is a broad range of behaviour, some of which is high risk, and a lot of the stuff that you just heard about is not illegal. Terrorism is high risk, high threat and very clearly illegal. There is a gap: local policing deal with the hate crime and I deal with that high end terrorist threat in however it forms. There is a gap in legislation and everything else to deal with extremism. That makes it very difficult for law enforcement, because we do not have the powers and levers to deal with that central space.

We are a much smaller operation. We rely on communities and victim focus. We are very much upstream of the counter-terrorism space. If we have intelligence, it feeds down, it does not feed back up. We are dealing with the implications on communities, the relationships with partners and the response that victims who have been targeted are reporting in, so we are perhaps more victim and community focused, I would suggest.

Can we focus on that gap? Can you give some practical examples of that gap and what legislation would fill it?

There are huge amounts of online content. You heard in the last session how, through algorithms, content is being pushed to individuals, including online gaming—for example, we have seen sites where the attack in Christchurch is now a game and you can go and commit the 51 murders online. That is not an illegal thing to do. Children are playing those games and that is creating a mindset. There are then a number of vulnerabilities that might play into how that behaviour plays out. The vast number of people absorbing this material will not go on to commit atrocities. The challenge for us how you intervene, how you identify who will, and where does it cross that threshold into terrorism. That is when it becomes an ideology and it then starts informing behaviour. Then you have capability issues and everything else. But there is nothing that allows us to intervene, from a legal perspective, in that extremist space, unless it becomes a hate crime or it tips a very high threshold into terrorism.

If I can give an example that you will be aware of it. When the “Punish a Muslim” letters were circulated in 2018, the perpetrator was investigated and arrested and dealt with by counter-terrorism colleagues. The impact on communities while that investigation was going on was huge. We were concentrating on levels of harm, fear of crime and the risk of people being promoted into criminality. We realised from forensic psychology advice that the perpetrator, from what we knew of them at the time, was not likely to carry out those threats. He was seen as a low risk of that. Our fear was that somebody motivated by the call for action would act out. He had been a perpetrator who had been posting material for a long time, but hit the jackpot when something got notoriety and got a response. The real fear is about dealing with the threat of fear of crime and the risk of somebody being motivated to act out. In the middle of that, we have something that you heard about earlier, foreign state bad actors, or foreign country bad actors, whether they are state or ideologically driven, who are trying to divide and destabilise our communities by, for instance, circulating historical photographs of acid attacks and claiming they are new attacks, and calling for people to take to the streets to defend their community, building on genuine fear to try to promote a breakdown of society. Whether that is a state-sponsored action or whether it is an individual ideological sponsored action, it is all playing out and affecting the community. That is probably why we fit together well.

Would making these acts of extremism illegal help, or would it drive it deeper into the dark web and make it harder for you to see what is happening?

It gives us a framework within which we can operate. We are definitely not asking for more terrorist legislation. We have a very good legislative framework that allows us to deal with those threats. This is about closing that gap between hate crime and terrorism. At the moment we have no levers at all. It opens up powers for us and other organisations, which will allow charity commissions and others to consider how they might intervene. But with no legislative framework, with no illegality of a lot of this material, that is an almost impossible thing to do.

AC Taylor, how does the police assess threats from people who have extremist views but do not have an ideological base? With what you responded to my colleague there, with your powers, my local police in my seat, the Metropolitan Police, said that they would not investigate hate crime. Are there particular groups that you can identify that are getting targeted and do you have the right balance of powers to deal with that?

Assessing threat is quite a complex thing to do. We take a number of vectors. We consider our intelligence basis; we work closely with the UK intelligence services, MI5 and others. We will do a comprehensive review of what threat an individual poses. When it comes to prevent, they are probably not at a threshold where they meet an investigative response, so they do not move into that Pursue pillar of activity. That is a multi-agency response, looking at all the things that you have heard about, from social background, some of the challenges that individual might have, healthcare and education as well as what we are seeing online. That will inform activity. They may pass into the Channel programme, if a clear ideology is identified, or we will pass on, predominantly to education and health, those individuals who we feel are not meeting a terrorist threshold, but do require a safeguarding intervention. One of the critical points here is in Lord Anderson’s report he talks about this big front door. We are in total agreement. There is a requirement for a multi-agency, well-designed, well-funded triage process at that front end. At the moment one of the challenges that we have with the capacity in the system and the way the system is set up is that Prevent is a thing that works and therefore it is an obvious thing to say that we will put them in Prevent. Prevent is set up to deal with terrorism; it is not set up to deal with all these other things. You need, underneath this initial triage, a number of hoppers that these people fall into that will provide the right intervention and the right safeguarding process, if we are talking about young people, to address that behaviour and stop them going on to committing atrocities. There is then a link to be made with us, within the Prevent programme and the interventions we make, or through MASHs and others with the safeguarding head or local policing for crime investigation if we feel some of the criminality that comes will meet a legal threshold but it does not meet the terrorist threshold.

With local authorities having been cut back with the MASHs, and the police as well, do you think that there is the capability, if anything was to change, to deal with it?

The capability is there, but I do not think the capacity is there. The system is not built to deal with this volume of demand. To give you a sense—I will provide you the actual figures, so don’t quote me on these figures precisely—we would see around 7,000 or 8,000 referrals a year. This year we are at 6,000 already. That is a 37% increase into Prevent. We are projecting more than 10,000 referrals this year. That is just into Prevent, which is one part of the system, and that part of the system cannot cope with that volume of referrals.

I fear I know the answer to this question after the last session, but I will ask it anyway. How effective is collaboration between the policing and the tech companies?

We have a thing called the CTIRU, which is the Counter-Terrorism Internet Referral Unit. We have a good relationship for the removal of illegal content. This is where you need to draw the differentiation. We have had some good results, but the numbers have come down. In 2022 we secured the removal of 4,750 pieces of content. That has come down and in 2023 was 930 and it is in the low hundreds now. We think that that is the case because the tech companies are getting better at removing illegal content. A lot of it does not meet the TACT threshold and there are some investigative reasons why we do not want to remove things immediately, because we need to keep those channels open. I think we have a reasonable relationship on what is illegal. The huge gap, as you heard previously, is where it is not. We cannot police the internet. It is simply too big and too difficult; we would never be able to do that.

Ours is a small operation. We do not trawl for hate. We never will send somebody out to search for offenders. We deal with complaints from members of the public. I would suggest, as you have probably heard, that the relationship has got significantly worse. We did work under a thing called a cyberhate working group that was established in 2011, and then with the EU as part of the high-level group on racism and xenophobia with the industry and built up quite useful structures through things like trusted flagger status for charity partners. We had a relatively robust relationship with police liaison teams. But because we do not have that power to issue a takedown notice, it is referring it against the platforms’ community standards. As you heard earlier, most of them have been reduced and some of them virtually scrapped. I have sat with people who have built up a relationship with, who sat and watched their screen go grey, or watched for their screen to go grey, to see if they have been made redundant. The loss of those resources has lost that ability, through things like trusted flagger status, to fly material to them for their consideration against the terms of use.

Is hate crime, which is illegal, proliferating on the internet?

Hate material is. Part of the process that we would do in the hub is take complaints from members of the public under most circumstances, and then make that assessment of whether the material is illegal or not. One of the issues is that most of our legislation is not extraterritorial, so by and large we have to prove the jurisdiction to prove the crime, even when we believe the material could be criminal. We talk about context being as important as content. In making that assessment, one of the big hurdles would be finding the jurisdiction. If it is on a social media platform, we do not see the IP address, and most of the platforms would not be willing to pass it without a court order—from a US court, in most instances.

If I could touch on volume into our Internet Referral Unit, since 2021 we have had 150% increase in referrals. We think that by 2029 we will have about 40,000 referrals a year. That is just stuff that is being referred by the public into us for us to look at. We saw a 48% increase after the attacks on 7 October, so these big events play into those referrals. We are also, as you heard earlier, seeing individuals and/or states using them as opportunities to spread further. After the Southport attack we saw a very significant increase in material, which was both misinformation and disinformation, and some very clear indicators that that was coming through robotics and other things rather than individuals posting online, and that is done in volume, often overseas where we are unable to deal with it.

Turning to working with stakeholders, how are you co-ordinating your efforts in working with local authorities, health and education to identify the key individuals who are carrying out the crimes and becoming radicalised?

Prevent is where we work. We have a very good multi-agency response within that Prevent framework. Where we are lending our counter-terrorism skills and our understanding of how extremism might play into terrorism is taking that network wider. We do a lot of work with the National Crime Agency around the comm networks and other things that you have heard of, trying to understand what information is available and how we might, as a broader law enforcement community, consider how we tackle them. Where we have to be really careful is that we deal predominantly with three core missions—counter-terrorism, counter-state threats and protective security. Our volume is increasing massively in all of them. We are seeing increased upstream threats from traditional counter-terrorism. We have had a 500% increase in counter-state threat demand—that is just what MI5 is giving us; there is more than that—and protective security and all the other challenges. We have to be clear that we do not want to completely muddy the waters on counter-terrorism, because then there is a genuine risk that we will miss those individuals who we really need to be focused on. I come back, I am afraid, to my gap. I think that the wider local authority support has the capability to do this, but at the moment the system is not set up to deal with this challenge. It is not funded, it does not have the capacity and in order to do this properly you have to do it as a whole-society approach—not just the system, not just public sector—to how we get our communications right. Parents are often the people who will see these risks in the first instance. We are seeing a lot of people involved in very small groups online. While you might absorb the material, that engagement is taking place in much smaller groups that are much harder for us to look at and understand. The very specific answer to your question is that we have a good relationship and we are working hard to understand these threats. The challenge at the moment comes back to this gap. I need to be careful not to go too far into that gap. Policing cannot go into that gap. Therefore we have this void where these behaviours and these challenges are promulgating.

While we focus on harm rather than individual perpetrators, we have good relationships with civil society who operate in this area at a national level. We have identified two gaps. One is the impact of international events on a diaspora living in the UK. If I could give one brief example, there was inter-religious violence in northern Nigeria in the run up to the end of 2024, where 300-plus people were killed in violence. It did not make a single newspaper, it did not make a single story on the TV here, yet we have 250,000 diaspora people with Nigerian heritage. So the tensions that exist within communities are capable of playing out here, and we do not have that. That international element and the impact on the UK living diaspora is one gap. The second gap is the point that was raised earlier about the reduction in the capacity at local authority level. It is undoubtedly the case, from the day where the CRE and the race equality councils were having that day-to-day reach. Professor Ted Cantle gave evidence to an extremism event fairly recently—he led on the response to the 2001 disorder—and talked about that issue, about the local authorities not having those structures embedded to feed the information in about what was happening. He talked particularly about Leicester and the violence and disorder from 2023, and how that was brewing locally, but not feeding into the national process. So there are two gaps, for me.

You have both mentioned challenges with resources there. Are early signs being missed with those young people, particularly within a school environment, say, who are being radicalised?

We have seen a significant uptick in referrals into Prevent. There is a question on the quality of some of those referrals. Some are absolutely right. Of course, we do not want to stop referrals being made, because we want the referrals to be made so that the professionals can make the assessment as to the right place for those to go. There is something about how we inform those referrals in the first place. We have some examples where you would laugh. I will give one: a kid at primary school who watches “Kung Fu Panda”, then goes into school the next day, kicks a pupil and is referred into Prevent. That sort of thing is not what we want, but it is important that we do not frighten people and avoid those coming in. We also want to be really clear that, if you are a young person who is viewing extremist stuff online, if you come into Prevent there is a risk that you get labelled as a terrorist when you are not; that is why I am clear that Prevent is not the sole answer to this. Are we missing things? Prevent is huge. It is difficult to say how effective it is, because you do not know what that individual would go on to do. We do know that we are very successful at preventing terrorist attacks and we have prevented a significant number in the last few years at a late stage. But, to quote that famous saying, we do not know what we do not know. However, the fact that we are preventing stuff would suggest we are picking up those signs from a terrorist perspective. There are wider risks, and we are seeing them playing out in society, from lower levels of violence to quite extreme levels of violence.

If I answer the first question first, what would support Prevent is that broader systemic response to this challenge. I met with Lord Anderson a few weeks ago and talked about this. He is not advocating that everything goes into Prevent. Prevent has a very specific role, defined within legislation, around the threshold for vulnerability of going into terrorism or susceptibility to radicalisation. To give you a sense, on average over the past years, 55% of our cases are closed as non-terrorism and they are referred on. Over half of those are referred into other organisations—32% into education, 22% into health, 18% into local authority and 13% into local policing. They are cases where we have said, “This is not a terrorist threat; we do not believe this person is vulnerable to terrorism. But we do believe there are interventions needed.” My concern at the moment is that the system is unable to fulfil that demand. What sits alongside that is, if you get this multi-agency effective triage system at the front end that is properly funded—funding and resourcing are always a challenge, and I absolutely understand that—with the appropriate interventions that sit beneath it, you create this system-wide approach—rather than what we always do, which is to talk about Prevent. Prevent is there for a very specific reason. That also means that Prevent can then focus on those things that it is there to do, which is to prevent people from becoming terrorists, in simple terms. Sorry, I forgot the second half of your question.

This comes back to the point where it is very difficult; we have no levers to do that because they are not committing any offences. We see what you might call sort of dog-whistles. Groups like 1488. You may have heard of it. It is a right-wing extremist group. The “14” refers to the white supremacist David Lane and “88” is the code for Heil Hitler. That in itself, 1488, is not an offence, and often the material is not either, so it falls into this broad category of legal material online. We have enough of a challenge dealing with illegal stuff, let alone trying to take on legal stuff. The answer to your question is that we can see it, but we do not have the levers or the tools to deal with it—that is law enforcement in its widest sense.

Can I go back to that again? I am struggling; this is a dilemma for liberal democracy, of course, dealing with extremist ideas that are not illegal, because you are allowed to have extremist ideas. How do we square that with a liberal democracy’s right to believe in strange and wrong things?

This is where this whole argument becomes challenging, and it is why successive Governments have been unable to define extremism. You have a right within law to hold extreme views and a freedom of expression under ECHR to vocalise those views, so long as you are not inciting violence and so long as you are not committing offences of terrorism. For me, looking at the scale of the challenge on the internet and those principles and those values we hold dearly in this country, we need to build an infrastructure, knowing that those things exist, that we can then deal with the consequences of them and properly identify those individuals who are vulnerable to this sort of material. If there are things that you can do to stop it being pushed to them rather than them looking for it, that would be enormously beneficial, and then a system that has the appropriate interventions so that you can tackle some of the misunderstanding and shaping of behaviour from this material. While it would be great to be able to control it, my pragmatic side says it is a big elephant and trying to eat it is challenging.

My sense is that in the early days of the internet we used to work with charities that wanted to do counter-narrative work and wanted to try to persuade people that a hostile and harmful view was not correct. We have lost that battle. That cannot happen because they lost the ability, through the algorithms, to get their voices heard in the same space, for the reasons we heard earlier. For me, the only possibility of doing that is at platform level through regulation and through self-regulation. I have a right to have extremist views and hold them; you do not have a duty to read them. Somebody is serving my views to you, rather than a positive version of the same narrative. It is that platform-level response, for me, that is the only solution left to us.

I am sitting here thinking about “1984” and George Orwell and his thought police. It is lovely to see some actual police here, but the problem is that there is a lot of bad thinking going on, isn’t there? There are young people who are effectively having malevolent thoughts put into their heads and we are not able to somehow control that. Surely the root of trying to deal with this lies in education. I am not sure that we are on to this enough, particularly in our schools. We need to be out there presenting that counter-narrative. I do not think that we should give up on it and say that we cannot defeat the algorithm. We ought to be out there trying to influence the way people think, young people in particular.

I completely agree with you. I am trying to say that this needs a whole-society approach. It would be very disappointing if we just said that it is too big a problem and we cannot do it. From a policing and law enforcement perspective, we can only deal within the law. We are not the moral police and it is not our job to question. We police without fear or favour and we will take that middle line in the best way that we can. But we must make every effort to do everything, from helping parents to spot the signs to understanding where vulnerabilities might exist and looking at education and the sort of behaviours that might be exposed. We did it with drugs. We were very good at explaining to people, “If your child is displaying these signs, you might want to consider this”. We need to be doing the same with this. We also need a system that people are happy to refer into that will not stigmatise people. The conversation on neurodivergence I thought was interesting. Yes, it might be one of the predicating factors for somebody being drawn in, but just because you are neurodivergent does not mean you are more likely to become a terrorist, and it is dangerous if we start labelling people in that sense. The more we can do through a whole-society, whole-system approach to this, educating and understanding, where we can push out the sort of things we are seeing at the extreme end, the things that are motivating hate crime, the more it will be hugely beneficial.

Can I add one more? One of the things that we have come to learn over the years from monitoring tensions and looking for reactions to incidents is that it is not always possible to follow a logic as to what will have the impact on society. If I give one example—the data was not complete—we saw signs of a 220% increase in anti-Muslim hate crime the week after the Christchurch atrocity took place, where the only role of Muslims was as victims of an atrocity as far away as it is possible to get. That was greater than we saw after Manchester Arena. My logic would have said that I do not condone retribution, but I understand it as a motive for humanity. But to feel that people are more likely to act out violently in empathy with a perpetrator because they have seen the perpetrator online has to be a change of thinking. Going back to that educational input about digital literacy, when I was educated I did not need that—I needed other skills instead. That has to be a fundamental part of how we equip young people to deal with this future that they are going into, and not the one I had to face.

I agree.

There was a report that may be of interest—you may well have seen it—on counter-terrorism policing and other counter-terrorist organisations across Five Eyes, “Young People and Violent Extremism: A Call for Collective Action”. That is about just that—looking at how can we get parents to be more engaged in their children’s digital lives, tech companies making platforms safer, some of the legislation that supports that gap. A lot of what you are talking about is set out in that report.

We have a final set of questions on what other measures we could use to address behaviour. We have some questions on youth diversion orders, which Peter Prinsley will lead on.

First of all, could you explain what is meant by a youth diversion order and how they are implemented?

Youth diversion is about finding effective ways to intervene at the earliest opportunity and deal with those people who are on a dangerous pathway. We would do the work to support them. They are about management of children and different approaches in the most effective way. They are non-criminal justice. We think that they are hugely positive in terms of an intervention that does not criminalise young people and does not label them as a terrorist or something else. The purpose is that early intervention, before prosecution, getting upstream and allowing us to try to tackle that behaviour before you go down a path that it is much harder to pull somebody back from. They are an early intervention, diversionary order that is about tackling behaviour pre-criminal justice.

Going back to what I was talking about in the last question, could you envisage curriculum content in schools that could be presented to help?

Yes, but I am not sure that that is the same as a youth diversion order.

No—it is just another tool, isn’t it?

Yes.

I am interested in how you take one of these 14 to 17-year-olds and you put them down a different track. We understood from the previous witnesses that these are incredibly impressionable young people and some of them just go the wrong way.

Yes. I disagreed with the gentleman who sat here saying that it was a surprise that 14 to 17-year-olds are more radicalised. I do not think that there is any surprise at all. At that formative stage of your life, you have no barometer to know whether something is true or not, you accept more of what you see, and you do not have the life experience to challenge. I do not think it is surprising at all and we would traditionally classify children as being more vulnerable. Yes, of course you need to educate and provide them with the tools so that they can manage their own safety as well online, some of which could be done through schools.

I go into schools and talk to young people about where they get their information from—I am usually talking about politics. They do not read newspapers or listen to the radio, but they all look at TikTok. If you ask them what they look at on TikTok, it is very short attention span things, often from people who have political views that are not exactly mainstream. We need to think about ways of teaching young people how to think—not necessarily what to think, but how to think.

Most of what you see on TikTok is not true.

The timing is unfortunate to be having this conversation, because in July there was a terms of reference agreement between the Home Secretary, or the Policing Minister, the College and then the PCC to do a review of non-crime hate incidents. That is not due to deliver until the end of December. My understanding of the statement from the Met Police side is that it will not investigate, but it will still record for intelligence purposes. That is one of the things that we will explore. There is a broad issue about non-crime hate incidents in general and their capacity. They emanated from the Stephen Lawrence inquiry and Lord Macpherson’s report into the tragic death of Stephen. They have had an existence and they sit across quite a complex area. They cover tragedies such as incidents running up to the tragedy of Fiona Pilkington and her daughter Frankie Hardwick, who both died in 2007. The review is looking at that broad issue and will report and it is fairly difficult for me to contemplate it before the evidence has come together. I have spoken to your officials and I am more than happy to feed that in before the end of the reporting period, but it is quite tricky at the moment. We have always recognised that those indicators, with interpretation, can give us the best barometer of community tensions. If we look at recent violence such as in Ballymena in Northern Ireland, which came from a trigger incident, the police and the partners there had been monitoring community tensions and knew that there was a febrile atmosphere, in the community. They could not have predicted the trigger incident, but being able to monitor those tensions as they arose gave them the best chance of being prepared for the violence when it did take place. It is an ongoing narrative, I guess, in policing.

As explained, we will still be recording, so we will get a sense and a feel for what is there. We have felt, both in terms of high-profile stuff and speaking to teams, that it comes a bit back to terrorism versus extremism. We need to focus on the things that are within our power and pose the greatest threat and risk. Non-crime hate does not always do that, and it puts our officers in an invidious position at times to be policing conversations and engagements on the internet, which is difficult. So we think that it is the right thing to do so long as we do not lose sight of what is happening, and that is why they are still being recorded for intelligence purposes, but we want to focus on those core priorities for policing.

Before we conclude, do you have any other things that you would like to say to us as a Committee that you do not think have been covered already?

One of the things that was explored some years ago under the APPG against antisemitism was the issue of what we referred to at the time of CrASBOs, that ability to control people’s behaviour. At the time the talk was about removing people from the internet. I think that most would accept that you cannot operate in modern society without access to the internet, but there are still things that could be done, such as controlling people’s actions by not having an anonymous persona on social media, not being on certain platforms or not carrying out certain behaviour. Having that as part of a criminal justice response might be a useful tool to have. There was a perpetrator who went to prison for three and a half years for inciting racial hatred in 2015, trying to arrange a march against what he called the Jewification of Europe, a march to Golders Green. While he was in prison, his material was still present online because we had no powers to take it down. Even though somebody had been convicted for illegal material, we did not have an authority over a platform to take it down and we did not have an authority to order him to do so. While there was an opportunity out of licence conditions in there, there are still things that we can explore that, on some of those persistent offenders, can have a real impact. There is another case fairly recently where somebody was suggesting that there was a threat to schoolchildren, and it was viewed 5.4 million times on X alone. I checked this morning and it is still there from September. Some of those non-traditional justice outcomes could still be valuable to us.

I remember from my time as a Minister the frustration of getting anything removed from the internet when it was not illegal. Child pornography is illegal in every bit, so it gets taken down, but nothing else. It was frustrating.

I will make a few points. Firstly, as I said right from the outset, Prevent is not a counter-extremism programme, it is a counter-terrorism programme, and conflating those two things is a risk. I align it to the Right Care, Right Person. We need the right people to be making the right interventions to prevent behaviour down the line. That might be Prevent, or it might be wider, which is the second point: this requires a long-term strategic direction, sustained investment, policy, development, funding, cross-sector collaboration that is not just focused on enforcement but is looking at getting upstream with education across the sector to intervene as early as possible and address those behaviours. Prevent relies on consent. If you do not consent, you do not go in. Things like youth diversion orders, where you can enforce some of this stuff, are a good way of doing that. Finally, this requires co-ordinated action across Government, stronger regulation, accountability for online platforms and closing that legal gap that exists for extremism. I am not asking for more TACT legislation, but closing that gap would be enormously beneficial for law enforcement and the wider system and it will provide those levers for them to actually do something about it.

Mr Giannasi said something that reminded me of something I had not asked you, Mr Taylor, about tech companies collaborating with you on illegal content. I assume all the tech companies are the same, but is that correct? Are some better than others?

I would probably need to come back to you with some detail. I do not have the detail on that, so I will come back.

If you could, that would be useful.

We do have a good relationship and we are seen as a good partner to work with them to remove that illegal stuff. I will get you some more detail.

Thank you very much. It has been a long session—not quite as long as it could have been, because we were not interrupted with the vote, so that has been very good and we are getting you out just about on time. Thank you once again for your comments and your responses to our questions. I will conclude the session.