- 18 minutes ago
Digital platforms have transformed how information circulates. Algorithms curate what billions of people see, often prioritizing content that reinforces existing beliefs, captures attention, and drives engagement. The result is a growing fragmentation of the public sphere. Instead of shared sources of information, audiences increasingly inhabit "echo chambers": personalized media environments where facts, narratives, and interpretations diverge dramatically. From rage-bait and viral misinformation to AI-generated content that spreads faster than it can be verified, the digital ecosystem is reshaping how societies produce and recognize knowledge. This panel explores whether common ground is still possible. When information is personalized and amplified by machines, how do we preserve shared standards of truth, scientific knowledge, and public debate?
Category
🤖
TechTranscript
00:19Hello everybody. Thanks for coming to see us today. My name is David Gilbert. I'm a journalist
00:26covering disinformation and online extremism for Wired. So for most of recorded history I think
00:35the problem that people have had is access to information. Obviously the internet and the
00:41world wide web changed that and today the challenge isn't so much accessing the information but
00:46knowing whether or not it's real. I think the access to low-cost AI generative tools
00:55has led to an explosion of synthetic content and we've got black box algorithms that are
01:03you know orchestrating our feeds which are leveraged by you know powerful often unknown forces who care
01:10more about control than truth. So I think the idea of this that we've had of a shared reality is
01:17quickly slipping from our grasp. So today that's what we want to talk about. We want to talk about
01:23those challenges, the real world impacts it's having and what potentially we can do to address
01:28these issues. So first can I turn to you Emmanuel. Can you introduce yourself quickly and maybe talk
01:36us through what we're actually talking about here. The state of the art of synthetic content as we know
01:42today in 2026. Absolutely. Good morning. I'm Emmanuel and I head investigations for Get Real Security.
01:50We work on real-time solutions to detect AI deep fakes and AI generated identity attacks. So both in real
01:59time but we also work on file-based detection. When we think about synthetic media sometimes the first
02:07thought is deep fakes it can be something maybe like Pope in a puffer jacket that happened two years
02:13ago but actually the state of the art is much more advanced now. We went from Pope in a puffer
02:19jacket
02:19to hyper realistic photo realistic images videos and voices with three seconds I can clone a voice.
02:27We actually prepared a video to show you guys to show you how they move they talk and they look
02:33like you
02:33and me so we can play it now. I know I look real you can see it the eyes the
02:39skin but I don't exist I
02:40don't exist but I don't exist they call us AI influencers because your brain thinks we're real
02:46but we are actually AI. These are AI influencers. Accounts like these are blowing up online they have tens
02:53of thousands of followers millions of views they build trust and they're not human. So these accounts
03:00come in all shapes and sizes and they're built to target different types of audiences. Millions have
03:05followed this so-called Buddhist monk but he's entirely AI and now he's selling digital courses
03:10and an e-book on healing. When someone cannot see what you give it is not because you are not
03:17giving
03:17enough. And this account pretends to be a real TSA agent sharing travel hacks when in reality it's an AI
03:24avatar created by a startup hoping to drive downloads towards a new app or the phone dies everybody's
03:31staring. And this influencer who seems to have attended the F1 race in Miami and even got to meet
03:36two Ferrari drivers? Entirely AI. That is not the real Lewis Hamilton or Charles Leclerc. And her account?
03:44It's driving sales towards an online clothing brand. Every single one of these accounts leads viewers to
03:50a store, a course, a product. They want to push viewers to buy. And most of these accounts are not
03:58clearly labeled as AI and when you take a quick look at the comments they tell us that most users
04:03believe these are real people. AI isn't just taking over music or even writing essays it's also entered
04:10the world of social media influencing. Two years ago when I covered this topic for ABC News, AI influencers
04:16were pretty obvious. They were static, easy to spot. They were a novelty built by creative agencies.
04:22Little Mikayla was the blueprint but her creators weren't trying to make her fully or overly human.
04:28And after that more personalities followed but you could always tell they were synthetic.
04:32Once this project comes to life one of my biggest dreams will have come true.
04:36Fast forward to 2026. I'm Isabella and today I'm here with my sponsor F1 Paddock Club.
04:42They talk. They move. They look like you and me. But you can even generate imperfections.
04:48All of this is now possible because of a new wave of hyper-realistic and powerful gen AI tools.
04:55And these all-in-one platforms that make it very easy for anyone to create and scale AI influencers.
05:01These aren't just experiments anymore. These are assets plugging directly into an existing
05:06commercial ecosystem worth hundreds of billions of dollars usually driven by human influencers.
05:12Until now. All you need today is one of these accounts, one of these influencers. Build an audience,
05:17post consistently, engage like a human, and over time users will believe. And the platforms,
05:22the ones built on the promise of authenticity and human connections, they don't seem to differentiate
05:28between human or synthetic on the platform. Engagement is engagement. And it doesn't stop here.
05:34Because if anyone can easily create hyper-realistic humans, a synthetic persona with an online history,
05:40a trusted following, they can use it anywhere. In a job interview, on a video call, inside of a company.
05:47This isn't just about content anymore. This is about an entirely new system that is being built.
05:52And one that is quietly reshaping what we perceive to be real.
05:59Well, thanks for that, Emmanuel. Just to clarify, you are real, right? You're not AI?
06:04That was not my clone.
06:06Right. Good to know. That was pretty shocking, I think, for a lot of people.
06:11Just quickly, you kind of spoke about it a bit in the video, but how easy is it to do
06:16this? Like,
06:17if someone here in the audience wanted to go away today and do it, how quickly and how easily? Like,
06:21do they need technical expertise to be able to create one of these avatars?
06:26No. And actually, as we were making that video, we had to update some of the footage because there was
06:31an update in model. And the platforms that you saw, those now include all different types of model,
06:37voice, images, videos. They're super accessible. You just put a credit card, you type in a prompt,
06:42and you can create these avatars, or you can create an image, or you can edit an image.
06:47And there are very little guardrails around what you can create.
06:53Massa, when we were chatting ahead of this panel, you used the phrase called ambient uncertainty,
06:59which kind of struck me as a really good way of explaining the kind of space we're in at the
07:05mode. Is it what we've just seen? Is that kind of what you're speaking about when you say ambient
07:09uncertainty?
07:11Yeah. I mean, Emmanuel's video was really great because it reminds me of so much of my time
07:16going on my Instagram feed and scrolling and myself being fooled. And I'm someone who works on this and
07:25is trying to follow all the latest trends. And it's becoming harder and harder to know what's real.
07:31And it really shouldn't be on us, necessarily, to be able to tell because the technology has
07:39accelerated at a pace that, you know, the guardrails have not kept up with. And this really does create
07:47a situation, especially when, you know, I work for Witness. We're a human rights organization. We
07:53follow things like conflicts, crises, any kind of situation where human rights abuses are occurring.
08:00And, you know, in these very high stakes moments where people need to know what the truth is. I mean,
08:06we were following the conflict in Iran, for example. And every single day, the amount of synthetic
08:14content that was coming out was being even debunked. Get Real, different organizations,
08:20organizations, including Witness's Deep Fake Rapid Response Force. We were contributing to this,
08:26but the lie travels much faster than the truth. And so this does create the situation of ambient
08:33uncertainty where you don't necessarily know what's true. The fakes are being passed off as real.
08:39And then the real, which I think is much more dangerous, gets perceived as fake. And within,
08:46you know, the work of human rights documentation, which has been always facing, you know, denial,
08:54if you go all the way back to like the early 2010s, where you were seeing kind of crimes in
09:00Syria,
09:01and you would see, you know, authoritarian governments like the Assad regime deny what was being documented
09:08in videos as being a lie. Now that's gone into overdrive, right? So it's a benefit to a lot of
09:16bad actors.
09:18Picking up on that, Philip, you obviously your company works to help organizations detect deep
09:26fakes as quickly as possible. Do you think we're at a point where the general public, or at least the
09:33kind
09:33of people who need to be aware of, like law enforcement media organizations have grasped
09:40this kind of idea that deep fakes and synthetic content is a huge issue? Do you think we've kind
09:47of got to that point yet, or we still got a bit of a way to go?
09:52Yeah, that's a kind of tough question to answer. Where do we actually stand right now?
09:59I think it's kind of mixed, I would say. So we see the awareness is rising. Also, as Marsha just
10:06said,
10:06for example, in the Iran conflict, Iran war, so much going on. And so many more people now
10:14understanding, hey, what I'm seeing online on social media, also, as we just saw in the video,
10:20doesn't have to be true. At the same time, we also see, like, if I watch at my own cousin,
10:27being like five, six years old, also seeing him scrolling through social media and assuming what
10:33he's seen there is real and not really questioning it. So I think this is also a huge topic when
10:40we
10:40talk about it, besides this whole, okay, on how to detect it, what we are mainly focusing on is this
10:46whole topic of awareness and building this awareness. I assume everyone attending probably
10:51also hear this panel, given this also title, that you are also familiar with the topic
10:58and the awareness is there. But at the same time, there are so many groups
11:02which are still struggling to understand and to grasp how realistic it can be.
11:07Because we still see reports of media organizations being duped by deepfakes and not whether it's
11:13just because they haven't got that muscle memory, maybe to kind of really check everything. And,
11:18you know, their first question isn't, is this real? It's kind of, let's get this up as quickly as
11:23possible. Do you think that one of the issues maybe that is that we're, and this is to any of
11:30the
11:30panelists, that because we're seeing so much, our brains and our perception is being rewired because
11:35we are seeing so much AI content now that ultimately, people just don't care that it's become the accepted reality?
11:49So, you mentioned the ambient uncertainty, which I have to credit to my brilliant colleague, Shireen.
11:54But one of the things we've noticed with ambient uncertainty is the fact that this confusion
12:00often leads people to just kind of withdraw and not want to have any critical analysis or not want to
12:06necessarily engage. So, there is this, I think, fear that we are moving towards kind of, I mean,
12:13the worst case scenario that a lot of people call is epistemic collapse, which is something we hope we
12:20don't get to where like we have no idea what is real and what is not. And I mean, I'm
12:25sure you're grappling
12:27with this too. Yeah. I mean, there's something deeply profound happening now that has taken it to
12:34the next level, right? Like we've, I've worked in verifying visual content for over a decade as an
12:41investigative journalist. And while deception is nothing new, and so our head of research, Tina,
12:47will say this on Saturday as well. Deception is not new. Manipulation isn't new. It's always existed.
12:53But the scale at which you can actually produce deception, the economy of deception has completely
13:02shifted. And then the hyper-realism, right? Because there was always a way that we could tell as humans
13:07and as trained experts, both in forensics or in visual investigations, there were ways that we could
13:14verify content. And when we talk about this, we're talking about visual evidence. We're talking about war
13:19crimes, injustices, any inkling of a breaking news in the last decade happened on social media. That's
13:26how we could prove what was happening around the world without having a journalist on the ground.
13:30That's being profoundly put in question because we have technology now that can replicate in a matter
13:36of seconds, real scenes. And without technology, we, and you know, governance and a whole bunch,
13:46not to get ahead of the questions here, but there's, there's going to, there is chaos already.
13:51There is. And I think we, we all saw that in the Iran conflict most recently. I think that,
13:57that for me, at least as journalists who's covered this stuff for a long time,
14:01the scale of the content that was posted on social media within hours of that beginning was
14:09just mind-boggling. Like, Philip, do you think that, like, how do you even begin to tackle that
14:16flood of disinformation? And because it's so easy to create now, and as Emmanuel said, it's so hard to,
14:23you can't know, you can no longer look at something and go, that's fake because the, the technology has moved
14:30on so quickly.
14:32Yeah, absolutely. So I think this is kind of like the big challenge. It is kind of a cat and
14:38mouse
14:38game. Those generators are getting more wheel and wheel. Like a couple of years ago, we had the
14:45chance to check, okay, like how many fingers do you have? Like, is there a sixth finger? By now I
14:50also saw,
14:51uh, like you can now buy on Amazon a sixth finger in order to try to, to bypass those, those
14:57kinds of
14:58checks. Um, so yeah, there is definitely a lot of, uh, creativity out there. Um, but yeah,
15:04also going onto this like whole setups of like social media where so much data is flooding in. Um,
15:11and it is now so realistic. We also see at the same time, even if it's not too realistic,
15:17it still has an effect on the human, on the consumer. So even if you have something flagged
15:23as this is an AI image, it can still have an effect on what you believe. Um, because like,
15:29it is a visual information and it travels so fast into your head and will have an impact there.
15:34Um, we're talking about kind of social media and the impact some people's brains were, you know,
15:39it's in abstract. Massa, can you talk a little bit about kind of the real world impact that this is
15:45having on people's lives, um, across the globe, especially kind of from a human rights perspective?
15:52Absolutely. Before I answer that though, I, I, there was something very interesting. You were
15:56talking about the media industry and I think Emmanuel was also talking about how kind of prepared
16:02everyone is. And I think we've seen so many examples of media getting it wrong. We're like,
16:08you know, uh, sharing a deep fake photo. I mean, there was one from the June, 2025 conflict.
16:14And I remember actually get real was one of the first to debunk the Evan prison strike video.
16:20But within the first few hours, this was headlines across sky news, BBC, it was everywhere.
16:26Um, so just, we didn't put it up. No, no, not you guys. You guys, they're, they're very good.
16:34That one was very tricky. That one was very tricky because it wasn't entirely synthetic.
16:39It was a real image. It's a real image that was from 2013. That was then animated using AI. So
16:46again,
16:47when you like, that's an extremely complex and layered piece of content to one detect. Yeah. And
16:54two, like an average person, of course, you're going to, it's a, it's a real, it's based on,
17:00I mean, it took me a while. I was suspicious because the context was suspicious, but like,
17:05we really did have to wait for like, get real that the first analysis, and then we had other
17:11forensics teams look at it. But the point is media institutions aren't prepared for this environment
17:17we're in. And all of these institutions that are supposed to be, you know, the places where we trust
17:24that are making decisions for us. And even places like we're talking about human rights accountability,
17:29like the court systems, they're not necessarily prepared for this new environment where digital
17:34evidence or deep fakes enters, or, you know, you have the, we've seen, I think, several examples
17:39in the US court system of people denying real evidence. They're like, oh, no, no, that's AI.
17:45That couldn't possibly be me committing that crime. So this is becoming an increasing problem.
17:50And these institutions, these traditional institutions haven't necessarily found the right means to
17:57tackle this. So that's just the main point I wanted to say in terms of, and actually, the first time
18:03I met Philip and Ditesia, because they were part of, you know, a UK kind of event, trying to understand
18:10how the UK government should be bolstering, and, you know, tackling this as it impacts like UK's democracy.
18:16So I was very happy to see cutting edge work like Philip's organization there too. But sorry, I realized
18:23it didn't answer your main question, which is the impact. But the impact of this is actually quite
18:28profound because we, I have seen it. I come from an Iranian background. So during the Iran conflict,
18:34I was following the work professionally and personally. And I was seeing things like, you know,
18:44the synthetic content was prevalent, but you had a context where people didn't necessarily believe the
18:53authorities in Iran, for good reason. They've done a lot wrong. And they've contributed synthetic
18:59content as well. But this doubt meant that when real documentation was happening, for example, there was
19:06a terrible tragedy, which was the US government bombed a school during the first day of the Iran war,
19:14back in February. And most people didn't believe this happened. Like people in my, in my own kind of
19:25networks, like for days, if not weeks, some until now, still don't believe the Minab bomb strike happened.
19:33And these are Iranians, some in Iran, because of this doubt, and immediately was accused of being AI. So
19:40many different pieces of documentation from that bombing was accused of being AI. And the real world
19:47repercussion is, in a context like that, people don't believe civilians are at risk from, you know,
19:54US or Israeli bombs. And like, there were direct stories of people not evacuating when they should have
20:01been, because the AI content was contributing to this information environment where people didn't
20:07necessarily know what was real and not real. And so there were real casualties of people who thought
20:14like this, who didn't think they were in danger.
20:16Um, Phil, like, because social media and the media itself, everything is about speed and whoever posts
20:25content first is rewarded. And, you know, that's, that's the imperative is get things up first,
20:31check later. When you're dealing in an environment like that, and you're, and lives are at risk, as Masha
20:37has pointed out, how do you tackle that from a technological point of view that you can combat that at
20:46speed or at scale? Or is it even possible at the moment?
20:51Yeah, so I think speed is super relevant here to react quickly. This is also what Masha and
20:57witnesses working on like this rapid deep fake response team, and you can quickly react to
21:03something and make clear like also we saw in the early example, this don't trust it. We see on a
21:11technical level, yes, it is possible to really have a fast detection. It is, of course, then also super
21:18relevant for the user to understand what are the reasons like, why is it flagged as AI? And we see
21:24also there. And then we also somewhat go into the the details of, of labeling, of showing from the
21:32generator side, hey, this is not a real content, which are super relevant, different approaches like
21:39C2PA, watermarking, everything on those lines, so that you don't even don't only have to check
21:46is it real, but then also the detection itself has to be super fast.
21:50But you're you're talking about solutions. But we we all know that people and the algorithm is
21:58going to feed people the most, you know, engaging content, the content that people are going to
22:03engage with. And so are like, how difficult is it to kind of put this back in a box once
22:12it's out
22:12there, for example, a video that shows a worker and that maybe a deep fake video and it's gone around
22:19the world and hundreds of millions of people have seen us, whatever percentage of them have believed
22:24this, none of those are really going to go and then fact check it and afterwards. So is is the
22:31biggest challenge, do you think, trying to stop the content getting out at source or or is it more about
22:38training people to to question what they're seeing?
22:43I think it's not either or. In the end, it has to be all of it. Like it is not
22:49sufficient to have,
22:50as you say, OK, as soon as something is online, damage is taken. Of course, then it is relevant to
22:57be as
22:57quick as possible in order to make sure not too much damage is taken. But at the same time, yes,
23:03definitely like it is not you have to stop it at the source, like on the generator level that you
23:09instantly directly from the one who generates a fake, it is labeled. But at the same time, you also
23:15need the awareness. But at the same time, this is also not sufficient. So you need also like on
23:21platforms where you can upload content also to have automated checks beforehand. Is this real or not?
23:28Because right now it is kind of like we as end consumers are faced with the challenge of
23:35can we trust what we see? And probably, let's be realistic, this won't change too soonish. But at
23:42the same time, it should go also into this direction. So that that's an interesting point. And it brings me
23:48on to something I want to talk about. You talk about, you know, the platforms and automatic detection.
23:53So that requires buy in from platforms. And most of these platforms are US based big tech companies.
24:01And over the course of the last 15 years, we've repeatedly seen them unable to really grasp
24:09issues around disinformation. Your colleague, Hanni Emanuel, who's meant to be here, but couldn't make
24:16it today, spoke recently about kind of that the tech has been weaponized against us and that tech
24:21companies want to burn it all down, which is a pretty pessimistic view of of the world, I think.
24:28So I'd like like, do you agree in terms of that tech companies are just not
24:34don't really necessarily care about their users and aren't doing enough? Or have you a more optimistic
24:40view of what's what's happening out there? I think if you hear Hanni speak, he's actually
24:47also an optimist, although because he wouldn't be doing this work of working tirelessly to find
24:53solutions, you know, with Tina and Adrian there and building a whole team of mathematicians and
25:00scientists who are tackling some of these extremely complex challenges. So if he were here, he'd say he is
25:07optimistic that we are able to find solutions. There are tech platforms who are working on detection,
25:17on labeling. Which ones? LinkedIn, for example, has the C2PA standard, which is think about it as like
25:26a nutrition label for media content. It displays the origins of a piece of content, how it's taken or
25:33created or edited and captures the journey. If we're talking about human rights and visual evidence,
25:41you know, that's something that's being talked about a lot. Do we attach a credential at the point
25:47of capture when we're capturing, you know, any evidence so that we can preserve that evidence
25:53throughout its journey? And that's a proof of that it's that it's an authentic piece of content.
25:59So you go on LinkedIn, you can see that, right? Is it across every single image? No. Is it a
26:05massive
26:05step forward? Yes. Do we want to see it across every single other platform? Absolutely. YouTube
26:12rolled out detection for deep fakes, impersonations. So a big issue, we're talking a lot about file base,
26:20but there's also real-time impersonations, voice impersonations. So YouTube creators and public
26:27figures were seeing themselves replicated and massive amounts of channels created with their
26:32own faces and voices making money off of their likeness. YouTube launched a detection where you
26:37can register your biometrics and it will scan and pull down anything that is not really you, right?
26:44So there has been some progress. And I think compared to the last 10 years of us having these
26:50conversations, they're actually happening at a much quicker pace than they have in the past,
26:54which makes me more optimistic. But they do need to happen across more platforms. And there needs to
27:00be sort of a... And sorry, just to add, there's also the Google Synth ID, which is an imperceptible
27:05watermark. So with any tool, any AI, any content created on a Google AI tool, there's this invisible
27:15watermark that you can detect if you have that technology. And now it's rolling out to open AI and 11
27:21labs.
27:22So there are some steps forward that are positive.
27:26I wanted to add, I believe in all of this optimism as well. We are kind of in this age
27:34where, you know,
27:35this is an age old problem. There's always been propaganda. There's always been disinformation.
27:39We're never going to solve this problem. But with kind of the accelerating pace of the technology
27:46technology and the sophistication of what we're seeing in audiovisual deepfakes or synthetic media,
27:53we're not going to have like one solution. I mean, we've been very invested as well within the
28:00Coalition for Content Provenance and Authenticity, C2PA, in creating the content credentials as being
28:06kind of one of the trust signals that we need in the world. It might not resolve everything, but it's
28:12definitely something that's needed. And of course, there's a lot of, I think, good teams and good
28:17intentions in a lot of the companies. And I hate to be the person at a tech conference to say
28:24this,
28:24but there also needs to be regulations that kind of demand this of the companies because there are
28:32really good initiatives. But at the end of the day, tech companies do hold a lot of power and
28:37responsibility for our information environment. And then, of course, how this reflects on life and
28:44death decisions in cases of conflict. And so we do need things like regulations. And there are some
28:50optimistic signs. For example, California has a AI Transparency Act. Parts of it are starting to be
28:57enforced this summer. And they are requiring things like content credentials as one of the open standards
29:05for provenance. And so we will be seeing how this will work in practice, hopefully, once that
29:11enforcement happens. So there are good signals like that where we need the buy-in from all these
29:17different stakeholders. There are. And not to belittle LinkedIn, but LinkedIn isn't the place where
29:24people go to find out information about breaking news. Just say you think it's lame. No, LinkedIn has its
29:32place, obviously. It's very important. But it's not where people go in a breaking news situation.
29:36They go to X. They go to Facebook. They go to TikTok. They go to Instagram. They go to WhatsApp.
29:41So they're the platforms. And I'm just wondering, like, Philip, from what you've heard from what the
29:46platforms are talking about or any interactions you've had with them, do you think those platforms that have
29:50typically kind of, especially in the last 18 months, kind of gone even more hands off in terms of
29:57moderation and disinformation detection on their platforms? Do you think those platforms are really
30:03invested in finding a solution to this issue?
30:09So from our perspective, from what we know, and this might change from a US perspective,
30:14I'm also curious to hear about that, is that there are teams within these companies looking into it.
30:21It is not always going with the approach of we have an image and we use, for example, an AI
30:29model to
30:30detect is it real or not. But there are also approaches which are going into this like content
30:35moderation approach. Who is posting this image? And with which other accounts is this account associated?
30:43How long does the account exist? How credible are they? And based on that, starting with kind of a
30:52ranking of whether you show this image to a lot of people or whether this could actually be somewhat
31:00weird because this account doesn't exist since a long time and is now posting so much content and
31:05it's going so viral in such a short amount of time. So we currently, from what we heard from some
31:13companies, they're also going into this direction.
31:17Amanda, do you think Mark Zuckerberg or Elon Musk who are ultimately in charge of these big platforms
31:24are really invested in trying to stop this, given so much of the content on their platforms now is
31:32AI and if it was removed or flagged or downgraded, that it would cut a huge amount of engagement and
31:40therefore cut their bottom line, which ultimately is what seems to be the most important thing for
31:46those companies.
31:49Yeah, I think that some of those platforms are attempting to rectify or have an attempt at
31:57transparency by labeling, right? But the problem is it's put on the creator to label their content.
32:04And for the time being, I mean, that is a real question that they've already answered.
32:10Synthetic content is allowed on the platform. So, you know, if it's not being stopped at the point of
32:16upload or like Philip mentioned, if there's no detection, because there could be a way to,
32:24for example, say, if you're uploading a piece of content that is relating to a war,
32:31let's put a detection mechanism there. That could be an option, right? But it's the same
32:38if you take it on... Is the technology there right now to do this?
32:41It could be, yes. I mean, yeah. And then you could also put it on the tools themselves.
32:50Why is it possible to generate this type of content? Why is it possible to generate... There are some
32:55platforms who have very strict safeguards about the type of content and people that you can generate
33:01using their platforms. And there are others who have zero restrictions.
33:06Marcia, did you want to end there?
33:08What Emmanuel was saying, it reminded, and you probably followed this, X had a specific policy
33:14related to conflict because the Iran war had just so much synthetic content that they rolled out the
33:22policy which said they would demonetize users that were posting AI content about conflict.
33:29Because obviously we know one of the reasons why there aren't that many guardrails to stop this is
33:35because it is profitable for companies. But the interesting thing about that is that there isn't
33:41very much transparency in exactly what kind of detection X is using. And we still don't know
33:48what kind of detection they're using. They're not part of C2PA, which is the Coalition for Content
33:52Providence and Authenticity. And their own AI generator doesn't necessarily have any, I don't think,
33:59a watermark that we know of. So, yeah. And I think that's one of the problems. We don't have enough
34:07transparency or insight or accountability from platforms like this.
34:11And just finally, Asha, as someone who's kind of on, you know, looking at it from a human rights
34:16perspective on a day-to-day basis, are you hopeful about what can be achieved in the future?
34:23I have to be hopeful or else why am I in my job? We're here fighting because we imagine, you
34:29know,
34:29a better world and a better system. I think there are signs to be hopeful. There are signs that indicate
34:35we should be working harder and getting more allies and more people to buy in.
34:42Okay. Ladies and gentlemen, thank you so much for listening. I just want to thank my panel
34:45who have been excellent. Please give them a round of applause. Thank you.
Comments