Good morning. Welcome to the Housing, Communities and Local Government Select Committee. My name is Florence Eshalomi and I am the Chair of the Committee. This is our second evidence session in our inquiry on modernising elections. I will ask my Committee colleagues to introduce themselves.

Good morning. Andrew Lewin, MP for Welwyn Hatfield.

Will Forster, MP for Woking.

Good morning. Lee Dillon, MP for Newbury.

Lewis Cocking, MP for Broxbourne.

I will ask our guests to introduce themselves.

I am Marianna Spring. I am the social media investigations correspondent at the BBC.

I am Sam Stockwell. I am a senior researcher at the Alan Turing Institute.

Thank you both for joining us. Obviously we just love elections, as parliamentarians, and many of us are commenting that we may have multiple by-elections coming up, but with elections comes a lot of stuff online, in print, misinformation and disinformation. The Government technically defines disinformation as the “deliberate creation and spreading of false and/or manipulated information that is intended to deceive and mislead people”. How prevalent is online misinformation and disinformation in recent UK elections?

It is hugely prevalent. Particularly over the past couple of years we have seen a spike, not least in AI-generated content and how convincing that has become, but also in the way that the algorithms, the recommendation systems on the social media sites, actively push misleading information or information without evidence. During the elections, the wild year that was 2024, where there were both the UK and the US elections, I spent a lot of time running what I called my undercover voters, which are fictional characters based on data research from the Pew Research Centre in the US and from some work we did with NatCen here. Those allowed me to see the feeds that were being pushed on different people. At that time there was quite a lot of disinformation on a local level, often about particular candidates. There was quite a high prevalence of hate and of conspiracy theory content, but since then I have found that the guardrails that had been put in place by quite a few of the social media companies have become significantly relaxed or lax, however you best like to describe them. That has resulted in the proliferation of some quite extreme content. Quite recently I did a documentary for BBC Two, which was called “Inside the Rage Machine”. I spoke to different insiders from all the major social media companies, who explained how platforms are built on outrage. That is not something new. I think we all know that now, but it feels quite important to talk about how these insiders say that the companies continue to actively make decisions that worsen the problem rather than improving it. That is very important when it comes to elections because it means that the outrage machine can be exploited by bad actors, whether abroad or in the UK. It also means that the companies themselves are not necessarily sticking to the commitments they have made to protect voters from misinformation, which could misinform their decisions on voting.

If the social media companies are actively looking at this, and you said in “Inside the Rage Machine” collating all of this, do you think it is possible to measure the impact that disinformation and misinformation is having on elections?

I think that one of the reasons why it has often been quite difficult to tackle this issue is because it can be intangible. There are two things that you can measure. The first is the prevalence of this kind of content. You can look at how many views and likes it is getting, but even particularly outrageous content will often be viewed by people who don’t necessarily believe it, but they are gripped or intrigued by it. You can look at the prevalence of this kind of content and then you can speak to individual voters—I spent quite a lot of time talking to different people—who have perhaps made decisions based on information that was misleading and they often felt duped or frustrated that they were not able to figure out what was true and what was not. I do a podcast series called “Top Comment” where every week we look at different posts popping up on social media. Most recently we have looked at the phenomenon called “decline porn”, which is content, often AI-generated but sometimes not, that makes certain places—say London or Liverpool or Birmingham—look like they are in decline in a way that they are not, contrary to the evidence. We have spoken to quite a few people who said that that kind of content affected how they voted in the local elections. It made them think, “I am very worried about the current political situation, so I am going to vote for someone else”. The problem is that that is a misinformed opinion. You can survey people, but it is often quite hard to draw the line between how people vote and the misinformation they see, not least because people don’t always recognise what they have seen might have been misleading.

Sam, one of the objectives of the Online Safety Bill was to tackle threats to democracy online and there is a lot of public concern around misinformation and disinformation during elections. From the work that you have done in your research, is that perception in line with reality?

In some ways I will say yes. When we cast back to a few years ago, particularly the concern about AI tools impacting elections, there was a lot of fear about the single, very sophisticated deep fake that would emerge a few days before the polls opened and would decimate a candidate’s reputation or completely undermine the integrity of an election. We found that in reality the level or volume of harmful AI content did not necessarily match those expectations. For example, during the 2024 general election we found just 16 examples across the whole campaign period where AI deep fakes and other forms of disinformation through AI tools went viral. Similarly, some other research that our colleagues have done found that at that time about 6% of UK voters who were surveyed in a poll could recall seeing a specific political deep fake during the election campaign.

Of that 6%, how many recognised that it was deep fake? I will be very honest, I saw a post and I had to get my staff to double check. They said, “No, this is fake. What is wrong with you?” The sophistication—it just looks so real.

Yes, the lines between reality and fiction are becoming incredibly blurred as these tools get more sophisticated. We found one of the bigger threats, which goes beyond just the election process and affects our democratic system more widely, echoes what Marianna said. In many cases we found that people were being confused in the comment sections on political content during the election, saying, “How do I know that this might be a deep fake?” Even when content was being debunked as a deep fake, people were unsure if they could believe it with their own eyes because of how realistic it was. There is concern about how this might erode trust in our wider democratic system.

Sam, you were obviously quoting the 2024 general election, but we have had a significant number of local elections since then. I get frustrated in this space because technology moves on so quickly. Examples of two years ago are out of date, but just happened to miss local elections and just happened last year. We are interested in: has it got worse since the 2024 general election? Was it 12% in 2025, for example? In the superset of elections we have just had, how significant is the problem now? Because that shows the trajectory of what it could be at the next general election.

I received some analysis from the Electoral Commission. It has recently done a deep fake detection pilot as part of the local elections, where it tried to monitor the online environment for political deep fakes that could be a cause of concern for the election process. It said that in a total of 900,000 posts that were analysed, it found 10 concerning examples that appeared to show signs of potentially having information in them that could, for example, damage the election process or people’s understanding of how to vote and things like that. Even though the volume is quite low, the problem is that they are very high-risk cases. Going back to the 2024 election, which is a very rich repository of evidence that we have, we saw that there were deep fake pornographic campaigns against female MPs. Obviously that is a cause of concern for the harm that that could do, with psychological damage to the individuals and potential chilling effects on others who may want to enter politics. We also found that deep fakes were causing violent comments against some of the MPs that were targeted. Again, there is the intimidation and harassment aspect. Obviously these don’t directly relate to the election results or outcomes, but I think we have to see that this threat is much more insidious and systemic. It is damaging our democratic process and norms more broadly outside of just the election cycle. Even though, as I say, it is low-volume content, it is very high impact in various ways.

I have a quick follow-up. If there is deep fake material going out against a specific candidate who represents a party in one constituency, obviously voters from across the country could see that and that could affect how they vote in their patch. It does not necessarily have to be related to one of the candidates that they are going to vote for. Does that make sense? How do you judge that?

One of the challenges that we have in election campaigns—coming back to what Marianna said—is that there might be a deep fake that someone sees that might go viral and millions of people view it and like it, but just because that happens, it does not necessarily translate into an actual change in voter attitudes or behaviours. One of the reasons is that people are exposed to such a volume of material over the course of a campaign. These processes last weeks, if not months, with the volume of material that people are constantly being exposed to. As our Turing colleagues have found, just 6% of people could recall seeing one of those deep fakes during the 2024 election. The wave or constant churn of information that we are exposed to means that just one deep fake alone is probably not enough. If we were to see that consistently being shared, consistently going out there into the information space, it might have more of an effect. I think we have to be quite cautious about conflating correlation with causation as well.

Yes. It is the drip, drip, drip of that, the point Sam was making, that can often end up having a much more corrosive impact, but not necessarily during the period when everyone is campaigning and people are thinking about the election. I have spoken to the Electoral Commission about this before. It poses complex issues for broadcasters because there are certain sets of rules during campaign periods and then there are not, but the content continues to exist on social media feeds all of the time. I did a “Panorama” that came out on Friday about AI and how it is used to spread misinformation. One of the issues I looked at was expanding on decline porn content, but specifically AI-generated videos that show, for example, people in the supermarket talking about halal meat or protestors attending what are styled as patriotic events. All of these videos were AI-generated. They were all highly convincing and they were getting quite a lot of traction around the local elections, but they had been shared for months and months and months. I traced some of the origins of this kind of content and quite a lot of it was being made by people who are based in Sri Lanka, Vietnam, the Maldives, miles and miles away from the UK. Some of the creators are people who were posting stuff because they think it is a bit absurd and satirical. Others seem to be just looking to get engagement and clicks and to exploit political outrage. Some genuinely seem to have a political motive, which is to leave an impression of what is going on in politics, even if it is separate from the reality and the evidence available to us. It is very important to understand the cumulative impact of all of these kinds of content. It is no longer about a single deep fake that misleads someone and then they make a decision about who to vote for. It is about the impression that this content leaves. If you look in the comments, there are some people who say, “I agree with this person” even if they are AI-generated, but there are quite a few people saying, “I know this is AI, but I agree with it and it echoes the sentiment that I have and I want to agree with”. It is the hyper-reality that we now exist in where people decide, “Well, because I agree with it, I want it to be true and I am going to make decisions on that basis”. That is the very worrying impact of all of this, and perhaps—aside from accountability for social media companies—social media literacy feels like one of the ways of combating that kind of issue.

Before I come on to sources, I think services like BBC Verify play a crucial role in helping to educate all of us about what to look out for and if there is the Gemini watermark. I can remember one video pointing that out, for example. Also, as the MP for Newbury, I want to put on record my thanks to the 77th Brigade, which exists to counter disinformation from foreign adversaries. That is an important service for the whole country, but based in my constituency. What can you tell us about the sources of online misinformation and disinformation during the election in the UK? Were they mainly foreign or domestic?

Now we have reached a point—and this is the new evolution of influence operations—where there is an eclectic mix of different people with different motives, who ultimately generate and cross-post and share the same content because it helps them to achieve their various goals, but they are not necessarily aligned in what they want. When I was doing “Panorama”, some pages I looked at were just engagement farming; they just want views and likes. They know they can exploit political outrage to build a platform and they can then sell merchandise or monetise the views and clicks that they get. There was one account where the creators were based in Europe and they said, “This is just part of putting out a particular political ideology that we want other people to buy into”. They were posting lots of anti-immigration content, again beyond legitimate criticisms or concerns that people might have. You also have the engagement nihilists, for want of a better word. Often younger people I have spoken to don’t seem to care about the way that the content they are creating could be repurposed for political aims. It is almost like being a troll in the original sense; they think it is kind of funny. There is a creator who made absurd videos of Croydon in south London, which had a waterpark and a zoo and all of this stuff. Lots of it was very absurd, like a polar bear and things like that, but some of it had taxpayer-funded labels on it and created an impression for local people. For many people I spoke to, that was quite concerning because they felt that it was reinforcing racist stereotypes and impacting what people felt about the area. I think that we no longer have a situation where foreign influence operations exist separate to the domestic home-grown content. In fact, the two things constantly interact. If you were running a foreign influence operation, you would look to co-opt movements that—

Domestic campaigns.

Yes, domestic campaigns that are having impact. We saw for a period that people would use real influencers. There are examples of where real influencers were posting content on behalf of hostile states or other groups supportive of hostile states. Now I think AI has enabled people to create very gripping, emotional, persuasive content incredibly easily and quickly, therefore that is new. If you think back to 2016, when we saw people running fake or bot accounts and posting divisive content, there was a period where that lulled because it was harder to do that because of algorithms and various other things. Now, once again you can make this content so easily and build a platform. The accounts I am talking about, the ones sharing the AI fakes, are getting millions of views for a single post and they are accounts that existed for say only six months.

This is a specific question to Sam. Some eastern European countries have been at the front-line of Russian attempts to interfere in elections; Moldova is a case in point. Are there any methods seen there that are of particular concern to UK elections?

Yes. We are starting to see foreign-linked networks, particularly from Russia, experimenting with AI tools to understand how they can improve the persuasiveness and the scale of their interference activities. One of the methods we are seeing more frequently now is the use of fake news articles and sources. Essentially they are using chatbots and generative AI tools, like image or video generators, to create what look like very realistic, authentic, authoritative news articles, often with narratives inserted in them, which are deliberately designed to be targeted at different audiences. Depending on the demographic in question, they will tailor them to different audiences. It is interesting though that I think we are still in a phase where they are experimenting and figuring out what seems to be most effective. Looking back about two years ago to the general election, we saw Russia utilise this tactic in the UK, targeting certain voter groups.

Did they use a particular platform as well? Was one platform predominantly used over any other?

I don’t think we have enough evidence on that. I think they generally try to gravitate towards the free available ones that have weaker safety restrictions and content moderation policies, as opposed to some of the more prominent commercial companies. It is interesting that we found that they sometimes struggle to understand how to use these tools effectively. To give you an example, in one of the articles that was trying to target a UK audience, they accidentally left in the prompt that they used to generate the article. It said something along the lines of, “Here are some things to keep in mind: Russia is good, Ukraine is bad” and that sort of thing.

Yes, and we would get preprinted leaflets that said, “Put candidate’s name here”.

Yes. They are probably learning how to rectify those types of issues, but it is still very much they are trying to see what sticks, what could be most effective. I worry that we will probably see more sophisticated efforts from foreign interference activity through the use of AI tools come, let’s say, the next UK general election.

Can I turn to the role played by algorithms? I am interested in both of your perspectives, perhaps starting with Marianna, on the extent to which algorithms are amplifying misinformation and disinformation, but also the extent to which they might be favouring one political agenda over another. I am keen to hear your overall view, but also any specific platforms where you have particular concerns.

When I did the “Inside the Rage Machine” documentary for BBC Two, I spoke to a whole range of different insiders from all of the major companies, particularly from Meta’s platforms, Facebook and Instagram, but also from X. I also spoke to some people who worked at YouTube and Google, specifically engineers that worked on algorithms, as well as people working on safety, and people from TikTok and some whistleblowers from TikTok. From speaking to them, what I found to be the case is that the algorithms themselves are designed to essentially ensure that the companies can get as many eyeballs on the content as possible to sell ads to make money. We know that now. Emotions are the most effective way of doing that, particularly negative emotions—outrage, anger, upset. Rather than the algorithms favouring a type of politics intentionally, any kind of politics that exploits or plays into those things is more likely to get the eyeballs and the likes and attention. We did an episode of the “Top Comment” podcast where we looked at basically which politicians are very good and effective on social media and those who are not. The content from Nigel Farage or Zack Polanski does very well because ultimately it often focuses on a single issue. It will be quite emotive and engaging and will often seem to be authentic or someone talking to the camera, whereas traditional political content tends to fare less well. We have some cringe posting, where it feels a bit out of kilter with what people are actually sharing. There is an interesting dynamic between the way that politicians now have to use algorithms to their advantage, but therefore that is affecting the trajectory and shape of politics because they are having to do so to get their content in front of people. That means essentially also playing the algorithms game. When it comes to what the companies themselves know about their algorithmic systems, one of the most concerning things when I was doing the “Rage Machine” investigation was how engineers would often say, “My job was just to optimise the algorithms, basically to get as many eyeballs on content”. They are saying A plus B equals C or X leads to Y leads to Z. They would say, “We are like the engine and the safety teams are the brakes. It is their job to put the brakes on when they feel that content could be harmful or cause an issue”. The problem is that the people who are the brakes teams were saying to me—including a very courageous whistleblower from TikTok—that they don’t feel that they are even able to keep up with the amount of content. They are often struggling to deal with the high volume, the content they are asking to prioritise or not prioritise. By the time they are dealing with it, it has often already gone viral and has had that impact. I hear time and time again from the people working in safety teams that they feel often that they are siloed or separate from the people working on optimising engagement. That is an issue for them. A lot of the companies have outwardly spoken about transparency as a solution to this, “We are going to be more transparent about our algorithms or our systems”. X recently released its source code, which essentially shows this point of outrage: arguing in the replies will get you more views, likes and engagement, and it doesn't shy away from that. It almost makes that part of its MO now. One of the insiders I spoke to was a guy called Brandon Silverman, who worked at Meta and ran a platform called CrowdTangle, which was an effective tool for looking at what is being shared and why. That was acquired by Facebook and then it was shut down by it. The bosses of all of these major companies continue to talk about transparency, freedom of expression and all of those things. A lot of the people from within the companies who care about transparency and freedom of expression worry that that is no longer a priority.

I am mindful of time. We have quite a number of questions to get through.

These are my words, not yours, but I would characterise that as algorithms favouring outrage. I am interested in both of your perspectives on whether there is any evidence, any cause for hope that some people are seeing this and turning away. I have seen some commentary in the Financial Times that we may have reached peak social media use, but I am interested in your perspectives, and tell me if we are wrong on that, as a naive hope. Also an important question: are there any platforms out there, even if they are small, that are actively trying to do things differently? If so, are they having any success in catching people’s interest?

From my perspective, it is uncertain where the trajectory could go. On the one hand we could see a situation where people start to migrate towards platforms that they feel align with their world views and that will have the type of content they want to see. There may be a risk of seeing a fragmentation of social media along political and ideological lines and a concern there. Alternatively, we could also see people move away from social media entirely to get certain types of news. Typically what we are seeing from Ofcom surveys is that unfortunately that is not the case, particularly for younger generations. They seem to be increasingly getting their news and political information from social media. A platform that has been quite interesting is one in Taiwan called LINE, which is the equivalent of Facebook. A few years ago they integrated an AI chatbot that was designed to allow people to share sources that they were unsure were legitimate or offensive or accurate. That would go through some sort of fact-checking review process, with a combination of AI and human review. Rather than telling the user if it was false information or not, they would simply give them a wide variety of sources to check out themselves, so it still gives the user autonomy over what they are looking at. It tries to give them more information about whether they probably should check out other sources. There is also some evidence from other countries that might be used here as ways forward to tackle these issues.

A very quick one, but as a point of hope, quite a lot of younger people I have spoken to for various documentaries and investigations, particularly teenagers and people in their 20s, have a real desire to interrogate stuff on their feeds. They don’t want to see a lot of this content. It is quite interesting that they often say they use the function on the platforms that says “show me less of this, show me less of this” and yet they are still getting that kind of content. Often what the algorithms think we want is not necessarily what we want. Anything that empowers particularly younger people and teenagers to be able to spot stuff that could be misleading and to interrogate it for themselves is a positive thing. In the “Panorama” I did, we spoke to someone called Professor Sander van der Linden, who works at Cambridge University. They have a game called the “Bad News” game, which essentially allows teenagers and kids to play like they are an online misinformation spreader and to learn more about the tactics being used to target them, the concept of prebunking rather than debunking. There are quite a lot of people working towards solutions. I always say my job is pointing out and uncovering all of the bad stuff and other people’s job is to come up with solutions, but there are people doing that work.

You have both outlined the serious nature of the problems that we are receiving on our social media accounts. What do you think would make the social media companies toughen up their approach to misinformation and disinformation? The Government are currently consulting on restricting or limiting social media use for under-16s. Do you think that would help counter disinformation and misinformation?

Again, with the caveat of investigative reporter, not a campaigner, all the usual stuff, I think that when it comes to protecting people on social media often a real gap exists between the speed at which legislation can move and does move, also sometimes the wider public understanding of new technologies, and then what is going on within the social media companies and particularly kids and teenagers’ experiences of the platforms. To some extent social media has become so integrated in all of our lives—adults or kids—that it is quite hard to put the genie back in the bottle in that sense. I hear time and time again from teenagers that they just want to be able to use the platforms. Most of the time they just want to talk to their friends and watch stuff that is funny. When I was doing the “Rage Machine” documentary, we went into a school where they are not allowed their phones during the school day and quite a few of them said to me, “I really hate seeing the AI content. I actually don’t like it. It really turns me off. I don’t want to see this kind of content”. From my perspective, I think that the best thing would be for people to listen to the users of the social media sites, particularly kids and teenagers, about what they want to see or not see. One of the solutions that has been proposed and quite a few experts talk about is having separate platforms that have different rules for under-16s, essentially, so you can still hang out with your friends and view content, but you are not inhabiting an adult space. That is often the inherent problem, that the algorithms push the same stuff to everybody. A 16-year-old will see the same things as a 40-year-old. We do not really do that in any other realm of life, but we seem to do that online. The other thing I would say is that when it comes to accountability for the social media companies, there are an increasing number of whistleblowers and insiders from the companies who are willing to speak out in a way that they were not even two or three years ago. I think listening to them in terms of solutions, features and work they have done has often been effective. There were people who were laid off at X who introduced new features to, for example, encourage people to think before they post an angry reply or to put safety mode on their accounts. A lot of the data showed that that was actually quite effective. Those features then disappeared. Anything that is receptive to actual design changes could make a difference. Sometimes we focus too much on what the companies look like now and then it all evolves and moves on and everything feels out of kilter.

For me, one of the concerns I have is around AI agents. Basically this is the use of autonomous AI systems that could carry out tasks independently of a human operator or user. Essentially you task this AI tool to go out and do something, and it can do it autonomously without the need for any human input. The concern in relation to disinformation here is that you could, let’s say as a disinformation operator from another country looking to interfere with UK politics, utilise a network of these agents to monitor social media platforms, understand what people are talking about around certain political parties or campaign issues, gather information on what is likely to be a very persuasive message and then basically churn out those narratives and do a constant evaluation as those messages are going out almost in real time. That acts as almost a feedback loop to ensure that your messages are going to be constantly adapting to local language, to cultural nuances, to different changes in the campaign and the election period. The concern is that we are maybe going to see static activities, but the use of AI tools to almost operate on their own to carry out these activities. That is a real concern, the ability for AI agents to go out there and potentially carry out these activities.

Yes, basically like waiting for the right moment, deploying your agents, getting them to do the evaluation, seeding the narratives at the right time, and then constantly changing and refining those messages over time. We are starting to see evidence of how that could happen in practice. It has not been realised yet, but I think that is the concern in the next three to five years.

Very quickly, I think that what we were talking about in terms of influence operations and the way that it becomes even easier to obfuscate and hide who you are and why and a lack of transparency means that often we might never know whether campaigns were being run by someone sitting in a basement in Saint Petersburg or someone in their bedroom in Burnley. It does not really matter any more. The impact is the same, but it makes it very hard to hold accountable the people who are doing this stuff. I think that the rise of AI-generated content that is very accessible, easy to make and very convincing is a real worry, not least because people start to disbelieve everything. My job is to investigate social media and I will see things and think, “Is that real; is that not real? I am not sure”. Often that content is not labelled. That could have a corrosive impact just on public trust in general. I think people will start to disbelieve everything rather than necessarily believing the misinformation. More generally, the social media companies are incredibly powerful and influential, particularly their algorithmic systems. Unless the issue around design is resolved, I think we will see a world where an increasing number of people—and it could be any of us—are effectively radicalised by their algorithms because they are seeing the same content, more extreme and more extreme. When we think of the next election, I am like, “What will people be primed to believe or disbelieve?” As a very quick example, there was a case, which you might have all seen, in Epsom where there was an allegation made about a rape occurring outside of a church. It turned out that that rape had not actually happened and there were some quite confusing communications between the police and the public. However, there were protests inspired by disinformation about who the perpetrators were, making allegations about immigrants that were just not true and disregarding the experiences of the person making the allegations. It felt like what happened with Southport again, and it does not feel like anything has changed to improve that situation. I would say the same around elections, essentially.

Obviously I work for the BBC, but I would say that it is having a public service broadcaster whose job it is to investigate this stuff and to do it for a range of audiences, to show people our workings, get to the bottom of the real-world consequences of this. It is why we do podcasts like “Top Comment”, because we want to be inside people’s social media feeds, effectively, and looking into the stuff that they are actually seeing. This also matters for older audiences as well, who get in touch all the time about the stuff they might be seeing on Facebook or Instagram. It is incredibly important that we have impartial investigative journalists who evolve in how we tackle this stuff, whether it is podcasts or documentaries or whatever it is, but do so in a way that resonates with the audience. I think that that social media literacy point, in the absence of action from the companies, is the only way of building up a societal resilience against this stuff, which I think most people across the political spectrum want.

A lot of this content falls into this harmful but legal bracket, which is the area that everyone has always had questions about tackling because of concerns around freedom of expression. What I always say to that is that a lot of this harmful but legal content actually inhibits the freedom of expression of the people who are targeted with it, whether it is disinformation or hate. What I would say, and this is from speaking to all the different insiders and whistleblowers from the companies, is that focus on design is crucial here. Anything that looks at the way the platforms are designed, whether that is with a view to protecting kids and teenagers or whether that is with a view to protecting adults, I think that is the right way of thinking about it. It is a principles thing rather than necessarily dealing with it at a moment in time and then everything moves on and we have not necessarily dealt with the issue.

Echoing Marianna, it is all about the system architecture and design. One of the things that I think we desperately need is a clearer way to know what content we are looking at. Is it real? Is it AI-generated? Where does it come from? How is it made? At the moment many people do not have an ability to know that and we are left in a situation where we are relying on comment sections or people to report on it. If we can get the platforms to provide that information at the point of content upload and ensure we have clear labels, we know where our content is coming from, who made it, I think that would give us a huge degree of extra transparency and awareness about whether we should be double-checking things, cross-referencing our sources and knowing what we should be trusting.

Thank you very much for that. As someone who triple and quadruple checks, if the social media companies could actually do their work, that would help us all. Thank you very much for coming before the Committee this morning. There was a lot of content there that we will make sure we review. Witnesses: Azzurra Moores, Phoebe Arnold, Hanno Fenech and Mitali Sud.

Welcome to the second part of our inquiry this morning on modernising elections. Can I ask our witnesses to introduce themselves, please?

I am Phoebe Arnold. I am the Policy Lead at Full Fact, the UK fact-checking charity.

I am Azzurra Moores. I am the Associate Director for Information Ecosystems at Demos, the UK’s cross-party think-tank.

I am Hanno Fenech. I am the UK Policy Manager at the Center for Countering Digital Hate, which is a research and advocacy organisation.

I am Mitali Sud. I am Senior Policy Manager at Internet Matters. We are a not-for-profit that exists to make the online world safer for young people.

Thank you for coming before the Committee this morning. You may be aware that the Foreign Affairs Select Committee recently highlighted a report and the fact that all the misinformation and disinformation we have been speaking about, seven separate Government Departments have responsibility for this crucial area. As a starting question to you all, do you think that this is an area that should be concentrated to one single Department? In a sense, if we are talking about a potential threat to democracy, should it not be taken more seriously?

I would say in addition to the Foreign Affairs Committee bringing this up, the Rycroft Review also pointed out that this is split. There is maybe what I see as a false dichotomy between disinformation and misinformation, like these things are running on the same systems. Yes, I think there is an argument for it. I would also say that the organisations involved in tackling disinformation in the UK are not transparent. Full Fact put in several FOI requests recently, basically asking which bodies are responsible and exercising decision-making authority. That was to MHCLG and DSIT. Both Departments, when they came back to us, said they would neither confirm nor deny any of our questions. I think there is a real transparency problem. If we want the public to trust that the Government have this under control, then there needs to be a lot more transparency and that could benefit from centralisation, for sure.

Would you argue then for a separate, standalone public body to tackle this?

Yes, and I think its job would be to monitor what is going on, to set up a national response framework, to communicate with other bodies, whether that is the police, social media platforms, anyone else who needs to be involved. It could do stress-testing programmes, evidence sharing and best practice.

I don’t disagree with Phoebe at all. What I would say to that though—and I am sure we will talk about this a lot in today’s session—is the way in which to tackle misinformation will require more than one Department. We do need the Department for Education, we do need DSIT, we do need MHCLG and the Foreign Affairs Committee, all these people who have such great insights to help us tackle misinformation. I think if it is left completely to one Department we are going to miss that multifaceted approach. I completely agree that it needs far more centralisation and it needs far more transparency, but we need all those experts in the room. We need all those Departments and probably we need a lot more resource than just what one Department can give us.

Yes, I agree that greater co-ordination is necessary. A specific example from CCDH research that we submitted to the Foreign Affairs Select Committee during the course of its investigation pointed out that an entity, Press TV—which had been banned by Ofcom from broadcasting in the United Kingdom because it was a propaganda arm of the Iranian state—was nevertheless communicating with a large number of UK users via its social media channels. This is just an example of how these things are not connected and rules and decisions taken in one area do not apply elsewhere.

I agree with everything all the panellists have said. The only thing I would add here from our perspective, because we do a lot of research with children and young people, is that what we are missing is an understanding that misinformation is part of online safety, which is part of electoral resilience, which is part of media literacy. Something I think we face at Internet Matters is that Ofcom, as the online safety regulator, seems to see its media literacy responsibilities as separate to its online safety. We would advocate for Government to play a joined-up approach across Departments as well as regulators.

Just quickly, I think there is a consensus on the value of a standalone body. My question is: can any of you cite any international examples of countries that have done this right and are further ahead than we are at the moment?

I don’t know of any that have a standalone body in the way that we are talking about. I do know that Canada has a public incident protocol specifically for elections, which is obviously relevant to this inquiry, which means that there is a predetermined way for Government to announce that there could have been election interference. I do not think it has been used, but even the fact that it is there goes some way towards restoring that public trust that I was talking about.

The Government are lowering the voting age to 16, so how can young people be supported to navigate that political mis- and disinformation online? Anyone can start, but I have a specific question for you, Phoebe.

I am very happy to kick off. Internet Matters conducted research with Full Fact earlier this year, where we conducted both quantitative and qualitative research with children. One of the biggest findings was that young people see support to help them identify false or misleading information as a shared responsibility. This is across parents and carers, schools, Government and social media companies, and Internet Matters wholly agrees with that. I will give three specific examples or three concrete actions, and I will be very concise. The first point that I want to land though is that this support needs to start at a very young age. We find that three-quarters of 13 to 17-year-olds, so about 78%, have seen content online relating to news, politics and current affairs, even children that are not supposed to be on platforms. The minimum age of social media platforms is 13. We know that a significant majority of nine to 12-year-olds are already seeing information online that could be misinformation or disinformation. We advocate for that support to start young. The first action that we recommend is around schools. I think Government have a big role to play in supporting schools with clear guidance, supporting teachers with teacher training, and providing practical resources for media and digital literacy. One of the reasons we advocate for this is because one of the findings from our report is that only about half of children feel confident in their media literacy skills. With Full Fact, we tested young people’s ability to distinguish fact from opinion online, their ability to spot information that is true versus false and their ability to recognise satire. Their own self-reported confidence was mixed, but then we also tested it with claims from the 2024 general election. We found that a lot of children said, “I don’t know” for some of the case studies that Full Fact had tested. We know that schools play a very big role, not just for children but also for parents. Parents are one of the main sources of online safety information that children go to and around 50% of parents say that they get their information about keeping their child safe online from schools. Schools play a big role in supporting both children and parents. The third action that I will highlight, which is one that I think the previous witnesses also discussed, is this idea of media literacy built into product design. Nearly half of young people think social media companies should take proactive steps to remove fake news, while over four in 10 say AI-generated content should be clearly flagged or labelled. For Internet Matters, what media literacy by design means is empowering young people particularly, but all users, to critically assess the information that they see online. This could include clear labels for AI-generated content, better context around sources, but even prompts to encourage users to think critically before sharing information.

Just to build on what Mitali was saying—

Sorry, just before you come in, obviously Full Fact said that teachers should be given practical support. Mitali outlined quite a few there. Are there any others that you want to bring to light for the Committee?

No, I think she has covered most of them.

She has ticked them all off, yes.

Yes, she is the expert. Just to build on what Mitali was saying, a lot of these platforms, especially during covid, did innovate, so they can do this. There were all kinds of different “read before you share” prompts on X. I remember that Google had a fantastic conspiracy theory redirect where it said, “There are not a lot of great sources for this. Try these sources instead”. That is just two examples. There were a lot more that have just been quietly dropped, particularly since the Trump Administration came in. One of the things that we have been recommending, and we tried to put forward as an amendment to the Representation of the People Bill, was putting a statutory duty on platforms to provide effective media literacy. That would have been definitely including by design, which, like Mitali said, is an effective approach that can also be evolved as new platforms and new capabilities come in. I would also say it is not just for social media platforms. I think the new way that Google is doing search is quite problematic. If design developments were made, for example, allowing people to toggle between different sources so that they were not seeing misinformation that had previously been fact checked—which we have seen appear in Google overviews—or being able to turn off AI completely, there are lots of things that companies like that could do to empower users and give them the context that they need in these quite radically new ways of seeing information.

Thank you. Are there any other comments?

Yes. I think one of the great ways to support young people online dealing with misinformation and disinformation, but many of the other threats they face, is for platforms to ensure that the tools that they provide young users for reporting and flagging information actually lead to an outcome. In CCDH’s research, too often we are testing whether or not the tools that are provided by the platforms for a variety of purposes, including these, lead to results for the users. In the vast majority of cases, we find that they are not performing effectively.

And that reporting back of the results as well. You have that cycle complete so you realise it was worthwhile reporting it in the first place.

It goes nowhere, yes.

Just leading on from that last question in terms of where it goes, you can see that prosecution for offences relating to the use of deep fakes around elections is very rare. I think about some of the deep fakes that we would have all seen used during elections. On that, what steps do you think should be taken to make sure that people are prosecuted a lot easier and quicker, in essence showing that it is not just about you reporting it, there are consequences to what you are putting out there?

There are many things to say. A low-hanging fruit here is that it has been the case that platforms have refused requests from law enforcement to identify individuals who have created deep fakes. That is a situation that makes any step that the Government takes to make this illegal much more challenging because then you cannot prosecute. This certainly needs to be addressed. More generally, the barrier to entry for utilising these tools for whatever purpose has been lowered so significantly in the recent past that CCDH has been advocating that the only means to address this is to put responsibility on the companies, the developers themselves. We have been attempting to do so in a variety of ways. One recent way was we worked with Baroness Kidron to amend the Crime and Policing Bill to make risk assessment duties on chatbots. Yes, not precisely deep fake creators, but a generative AI tool. Again, utilising the criminal law is of course a narrow path that does not address all the problems that we see and that CCDH has identified, but we do need to be taking these steps and quickly to address these harms.

One of the things I argued in the last Parliament during the Online Safety Act was around the fact that Government put forward proposals in terms of fines for the social media companies. A number of us cross-party were saying unless the directors and chief executives were personally liable, fines would be peanuts for them, essentially. Do you think that is something, actually finding someone who is going to take responsibility for some of the content produced on their various platforms?

Yes. CCDH has supported senior manager liability. The reference point here is that this did look like it was one of the things that made a difference after the global financial crisis, in that having a named individual, a senior manager, liable for the consequences of a service or an organisation is effective.

Yes. The scale of the problem is great and has got significantly greater. In the earlier panel, Lewis asked about the changes that have occurred since the last general election in 2024. Research from CCDH has shown that things have gotten much worse in terms of the level of abuse and the prevalence of abuse directed against candidates since that election. Our hypothesis that we have been testing has been that this is owing to decisions by major platforms to weaken their enforcement policies, so not necessarily to change what they say is not permissible, but rather to change how they enforce what they say is not permissible. I have the specific findings here, but in effect we gathered about 8 million comments directed at politicians and ran an AI classifier over them to identify instances of abuse. In comparing the six months prior to Meta’s policy change announcement in January 2025 with the six months after Meta’s policy announcement change in January 2025, we saw a trebling of abusive comments directed at politicians, indicating to us that the change here—the thing that has made things worse—has been an independent decision on how they enforce on their platform.

Sorry, did you analyse whether those attacks against politicians were against certain types of politicians?

Yes.

Were there more attacks against progressive politicians than others, for example?

No. We looked at the prevalence overall, but it included politicians from all across the political spectrum and across institutions.

Yes. We are seeing that abuse directed at candidates and politicians is directed at them all.

CCDH worked with partners in the Online Safety Act network to publish an elections code of practice in February of 2026. That is a good place to start in addressing the protections that need to be offered to candidates. Its specific proposals include ensuring that the tools candidates can use to flag abuse actually lead to an outcome from the platform. It would also like to suggest a risk assessment on platforms during election periods or around election-related risks that can be overseen by the regulator, Ofcom.

I completely agree with everything Hanno was saying, and I think the code he is suggesting is exactly what is needed. The only thing I would caution is just to make sure we are not going back to questions around the scale of the problem. We know how bad it is. The Electoral Commission has conducted its own data, and of all the candidates surveyed at the last general election, 70% said they received abuse. Of those 70%, 65% said that came from online abuse. There is a real concern at the moment, particularly with legislation that is passing through the Representation of the People Bill, that we are not recognising—exactly as you said—the prevalence of online harassment, but also the fact that what happens online really does transfer to offline as well. It is just cautioning the fact that we have a real moment to make a difference, but the Government are acting incredibly slowly and not recognising the scale of the problem.

For example, rejecting the amendment that would have brought this code into law. The Public Bill Office rejected it.

I think they are not incentivised not to. One of the things that Full Fact has been talking about for a while now is bringing certain types of misinformation risk assessment within the purview of the Online Safety Bill and therefore putting it into Ofcom’s codes, which would require these companies to assess and mitigate things like risks to civic discourse, risks to elections, potentially an element of risks during crises, but this has been kept well at arm’s length by the Government. Obviously Ofcom cannot take any action on that if it does not have the power to. Until we accept that certain types of misinformation are harmful, i.e. people will believe them and they may act on them—and that may be limited to specific circumstances like the ones I just outlined—I think until we all agree that those are real risks that may be legal but harmful, like Marianna was talking about, then there is going to be no incentive for them to do anything. The other thing I would say is we have all the evidence now. We know that when the companies have cover to act or not act, they will stop. We have seen that over the last six months, quietly dropping stuff, publicly dropping stuff.

I would add to that that if the Government cannot decide how to tackle misinformation, I am not sure that should be a challenge just left to the social media companies. We need the Government to be taking that action first and recognising the problem. Going back to the first question on why we have so many Departments that are involved in tackling the issue of misinformation, I think it shows, first, there is no centralised responsibility, and secondly that there is no centralised understanding of how big and concerning the problem is. That is not something we should leave to the social media companies. We need, as a country, to have a Government that leads that charge, to say what the problem is and then put those obligations on social media companies. That needs to happen first, in my opinion.

The only thing I would add to that is based on research we conducted on children’s news consumption. One of the things we found is that children are presented information side by side that is very difficult for them to distinguish. This is a result of how social media platforms are designed. Just to give a stat to help contextualise this, of young people who consume news on social media, only 40% of them get it from news outlets. The same amount, at 40%, would get it from influencers, content creators or even friends’ or families’ accounts. This information, because it is presented side by side, is very difficult for young people to understand what is true or false. As Azzurra said, this is intentionally designed. Platforms are designed in such a way. I think about 40% or four in 10 young people who get their news do not follow news-focused accounts. It is because of how recommender systems are designed. It is about how content is presented that makes it difficult for children. As Azzurra pointed out, this is an intentional design mechanism. It is an intentional design choice by platforms, and I do not think we can rely on platforms to correct this themselves.

First, I am not sure any of your solutions are radical enough to try to deal with the scale of the problem. That is just a point. I think you need to go away and think about what more needs to be done. I get that they can take off 1% and all the rest of it, but I have yet to hear anything radical where I think that is a game changer and will solve a lot of the problems we face. Why do you think the Government are not taking action on this? Do you think there is no political will? We have mentioned a number of times that there are lots of different Departments responsible. The Prime Minister could come out and say, “Right, this is the Minister responsible and this is where it is all going” and this could move forward with that. Why do you think they are not doing some of the simple things to make the small steps to move in the right direction? Why do you think that is, just lack of political will?

I would politely push back on the lack of radical solutions, if I may. I think there is always a need for twofold solutions. On the idea of what we can do to tackle existing legislation—and that is something that we at Demos and I know others on the panel agree with—we can start to do that through the Representation of the People Bill, and I am happy to outline that. Then I think there are more radical solutions, but tackling something like deep fakes or AI-generated misinformation has huge risks to free speech and we need to be careful in how we approach that. We know that the Government are concerned about this issue. In fact, when publishing the election statement last summer, Keir Starmer himself said he understands the threat that misinformation plays on democracy. I do not think it is a lack of political will. I think it is a mix of a lack of solutions, a fearfulness, and platforms that are so powerful and perhaps do not want to do anything. There is a boldness that is needed. Certainly at Demos we have put down a number of amendments via the Representation of the People Bill that would start to do incremental changes to tackle misinformation. That would be things like clarifying existing law that could tackle deep fakes, things like creating a centralised ad repository that would give the public the power to look at various political ads and understand themselves what is real and what is fake. There are loads of other suggestions we can talk about shortly, but what we would like to see is the Government recognising the problem first, which we think they have done, and starting to tackle the problem through existing legislation. Big wholesale reforms are needed. Do we need AI legislation? Absolutely, but that is not something that is going to happen quickly. To me, the scale of the problem is something that needs to happen via amending existing legislation so we can see that in the next year or two years.

What is the radical solution? By the way, I did not say whether I agree or disagree with any of the solutions or radical solutions. I completely understand that it is a finely balanced act to try to deal with some of this stuff, as you quite rightly pointed out, about free speech. What is a radical solution then, in your eyes?

I think, for example, in election periods we would like to see far more regulation. I do not think it is that radical, because we know that the public massively agrees with that. In polling that we conducted over the local and devolved elections, over 50% of the general public want to see more regulation during election periods. Election periods are pretty unregulated in terms of the information that exists online. You, as a political candidate, can share a lot of information and no one is going to say anything about that. We would like to see radical things like punishing chatbots, saying that AI companies cannot just allow chatbots to spread misinformation during election periods; during our regulated election periods, you have to use accuracy; you have to ensure that chatbots do not contain certain amounts of bias; you have to do things like publishing more transparent results about what is happening behind the scenes on platform data. That is going to help us start to find more solutions. These platforms currently are a real black box, so we do not know what is happening behind the scenes. If we don’t know what is happening behind the scenes, we cannot find solutions.

I agree with everything Azzurra said, especially about freedom of speech and incremental solutions perhaps being a way to protect that. I also think there is an element of multilayered solutions for a multifaceted problem, which I am sure you are aware of. Even within something like content provenance, it is not as simple as legislating, for example, that camera manufacturers and social media companies have to have a watermark when a picture is taken or uploaded online. I think that you have to look at each stage of the process and work out what is going to give people the context they need in order to avoid those watermarks falling off, because sometimes they will get cropped. I guess I am saying there is a pipeline to information being produced, being disseminated and being consumed and then maybe reshared. I also think there is an element, maybe as Azzurra was alluding to, that these are just big, powerful companies and the UK has set a lot of store in AI data centres and growth in AI, so you can see why this puts the Government in a tricky position and why maybe there is no legislation for AI in the King’s Speech.

Essentially, we cannot allow the social media companies to do their own homework because they are failing on that. If you look at a deep fake, once it is out there, it is there. If the candidate who was subject to that thinks about standing in the future and that comes up, that internet footprint or that digital mark is there and it is quite hard to clamp down on. In terms of my colleague Andrew Cooper’s earlier question, should we not be looking at focusing on more outside of election time? The big focus seems to be around election times, rightly so, but the damage for a number of different candidates, people who are thinking of standing, serving politicians, is sometimes done outside of election time.

I completely agree. I think we are now existing in a political system where we are permanently in a bit of an election campaign. Whether we are having general elections or by-elections, it will not be a shock to everyone sitting around this table that we are just constantly debating which party should be in power. Absolutely, that should change the way in which we look at regulated periods, because if it is constantly an election period, then we need to have better safeguards in place. I wanted to go back to your point around the platforms not doing anything, because there was some interesting data in the previous session that Sam Stockwell talked about, which the Electoral Commission shared with us prior to this session. It shared an interesting point, that it went specifically to X and an account on X and said, “Can you please remove this one account that is spreading a huge amount of misinformation about election periods?” and X refused. X refused because it said that it did not contravene its internal policies. I think that is a very good example. X in that example did nothing wrong, but what we then need is the Government to come in and say, “No, X, you have to remove this. We are going to legislate to say that if an account is spreading false election information, then that is not good enough”. I think you are totally right. We cannot leave it to the platforms to mark their own homework. We have an Electoral Commission that is pushing them and pushing them hard and they are choosing not to do anything. They are currently well within their right to do that.

I wanted to ask about online political adverts. To take you into my world as a Back Bench MP, I think they have an important role, £100 here on a campaign to reduce the speeding limit or £250 there on a petition for a new GP surgery. My experience has been that the amount of abuse and comments on targeted ads is far less than a mainstream post because they are not seen by people a couple of hundred miles away who might be bad faith actors, so I think they have a role to play. However, my question is on the case for an online library so that members of the public can see who is advertising to them and what agenda. I am interested in your views as a panel on the case for a library and whether you support it.

I certainly don’t disagree with you that targeted ads are a huge facet of our political system and we need them. We need to be able to tell people in individual areas what a certain MP is going to do for them. To me, the case for a centralised political ad library is really centring around the idea of transparency. It is allowing us to understand who is paying for political adverts and where they are targeting. If you are a voter, it is giving you the ability to say, “Hold on, this ad looks fake. I am going to check it myself”. It is empowering voters to do that. Demos had put down an amendment with Emily Darlington MP at the Committee Stage of the Representation of the People Bill to put forward a centralised ad repository. We think this would be for paid-for only political ads, so it helps distinguish that question of what is a political ad. We think it should be paid-for only ads. Our whole argument here was around transparency. We need a system where we give voters the ability to check things for themselves and to allow politicians to check things themselves. In the last session, we were all saying it is hard to know what is real and what is fake. If we cannot legislate, let’s at least give people the tools and the ability to do some of that homework themselves.

Yes, and there has been a bit of pushback, I think, in the Committee for the Representation of the People Bill about the risk of advertising platforms pulling out of the system, which Who Targets Me has done some good work on. One of the things it pointed to was the risk that if you remove political adverts from the online environment, you then end up with a situation where MPs who have a powerful social media presence do very well and then ones that do not—maybe they do not have the funding for that or they want to talk about policy more—cannot reach those people, as you just said, who they are trying to target. You also risk a situation where perhaps political speech ends up becoming more extreme to try to game the algorithms. Some of the things that it and others have come up with to avoid pull-outs of Meta and others in the EU are co-designing the libraries with the companies, within reason, forcing a review of the legislation to see how it is working in practice, introducing penalties for companies that do pull out and stop running political ads, and then finally making it a condition of advertising in the UK at all. There are lots of ideas there for how to avoid that situation.

I don’t have much more to add, but CCDH has long supported political ad libraries as one of many of the transparency measures that we have tried to place upon platforms. Another one that is pretty critical in this arena is the final provision of statutory data access for researchers like the organisations here. This has been long promised and long delayed, despite there being enabling powers delivered via the Data (Use and Access) Bill last year.

We have talked in this panel and the previous panel about there being very little incentive for social media companies to deprioritise misinformation and disinformation because they are trying to get as many eyes on ads to generate their revenue. How reasonable is it, via legislation or Electoral Commission code of practice, for us to require them to remove any disinformation or misinformation and how quickly can we do that?

I would start with an overarching line, which is the Government have the ability to dictate the terms and services of how platforms operate in this country, and I think sometimes we forget that. I think sometimes we think these are huge, enormous corporations and we cannot do anything. We need to reclaim that understanding that we can have a say on how that is done. As long as we do it fairly, in consultation, usually with the general public’s support, we do have the ability as a country to make sure our citizens are safer. How we do that should be up to the Government to design. To me, that would be the overarching point I would like to just raise.

The panel of experts says there is no legal or technical reason why we cannot make social media companies tackle misinformation and disinformation.

They have to tackle it in the EU. There is a model already.

Do you think the transparency approach, whether that be advert library or other things, is helpful but is not the full solution? That is good. Then I am going to ask: the Government have the legal ability to do it; they have the technical ability to do it. Why have they not done it?

There is a long history of that fight, particularly around minimum standards for what social media companies have to and do not have to do. If I recall the political arguments at the time, there was a difference between saying you have to have a system and process that addresses this risk and saying, “You, company, have to do these things”. It was thought that that was a speech issue and that you were dictating to the companies what was permissible on their content. There were political fears about that, but I cannot reconstruct those arguments fully at the moment.

I get the freedom of expression argument, but it is not freedom of expression to stop someone shouting fire and abuse at a remembrance service, and that is effectively what we have. No one thinks that is not freedom of expression. I am going to probe more on why the Government are doing it. We are hearing that disproportionately it is foreign actors and people very unhappy in this country trying to make this country think we are worse than we are. The Government have a political incentive to do it internally. Is it concern about Trump and America and Musk leaning back on the UK? Do you think that is one of the reasons why they are not doing this?

I certainly would not want to speculate why the Government are not doing it. My own assumption is that we have spent the last seven years having a debate around how to protect children online, and that is a debate that is not finished yet. Until that is fully resolved and we see legislation fully implemented, I think we are not ready to have that conversation yet. To me, that is incredibly concerning when we are talking about election periods, because what is going to happen is we are going to have a big crisis. We are going to have something that impacts the results of an election and that is going to force us to expedite legislation quickly. To me, it is where our focus has been on children, and perhaps rightly so. We need to protect our children online, but we also need to think about our election periods, which are particularly volatile at this current moment and will only be more volatile in the 2029 general election.

I would completely agree with you. I think what we are seeing, if we are not seeing one major crisis, is a massive impact on public trust. Marianna Spring mentioned it in the last panel. That really is the result of a lack of action, in that the public do not know what to trust and they do not know who to trust. That is the biggest result and impact we can see on a general election. What I meant more was one major incident, one deep fake that would really rock an election result, whether that be nationally or in a constituency. I think we are almost waiting for something like that to legislate more quickly. However, I agree with you. I think we are already in this real democratic crisis where people do not know what to believe.

Yes, and just to add quickly to that, we recently did some polling, where 80% of people said they were very concerned about political misinformation. The thing that caught my eye is that 33% say it affected their likelihood of voting in an election. Other people have done polling on this as well. Mitali has done polling with us among children. I do think there is a link between misinformation and participation that needs exploring.

To your point on radical solutions, as an organisation that focuses on young people, media literacy is one of the ways that we look at tackling this. There are countries in the world that look at media literacy like a national security challenge. It is related to the functioning of their democracy and this is something that Internet Matters strongly advocates for. To Phoebe’s point, in our research we found that six in 10 young people say they ignore what politicians and political parties say because they do not know if they can trust them. We start to see this real concern about young people, even if the franchise is lowered, not having an interest in using their ability to vote because they cannot trust the information environment. I would point to some countries that maybe have a more material risk of foreign interference than the UK does. Looking at media literacy like a national security challenge is something we would advocate for.

This has been an interesting and useful session and it will help us in our deliberations and questions, making sure we look at outcomes, as my colleague Lewis said, for the Government to take up on this. As Andrew Cooper highlighted, the longer we leave this and that drip, drip effect, it will impact democracy. For us all, it is around defending democracy, which is vital and important. Thank you for coming before the Committee this morning.