Seeing Is No Longer Believing: AI, Consent and Trust in Visual Media

The transcription explores how artificial intelligence in visual media challenges ethical concepts of consent and trust. It discusses the blurred boundaries of consent in AI-generated content, highlighting issues of control and awareness for actors. Trust is also examined, focusing on the impact of AI on relationships between actors, studios, audiences, and societal trust in visual media. The text delves into the implications of AI-generated content on authenticity, performance, and societal trust, emphasizing the need for evolving safeguards and frameworks to keep pace with AI advancements. The ethical dilemma lies in maintaining meaningful control over one's likeness and ensuring audience trust amidst the rapid evolution of AI in visual media. Over the past few years, artificial intelligence has moved from generating quirky images to producing hyper-realistic visual media. Today, AI can recreate deceitful actors, digitally replicate performance and generate cinematic quality video from text prompts. But as AI becomes embedded in visual media, from deepfakes to AI-generated films, the ethical question is no longer simply, is this impressive? Instead, we need to ask, what happens to consent, and what happens to trust? In this episode, I explore how AI in visual media challenges two core ethical concepts, consent and trust. In the Black Mirror episode, Joan is Awful, a woman discovers that a streaming platform has created a television show about her life using her digital likeness, which is something she technically consented to through terms and conditions. The episode feels dystopian, but it mirrors a real direction in media technology. AI systems can now replicate faces, voices and performances with striking realism. So while the show is fiction, the ethical questions are not, and are surely, in place today. Let's begin now with consent. Consent, ethically speaking, must be freely given, specific, informed and unambiguous. In traditional filmmaking, an actor consents to perform in a particular project, in a defined role under agreed conditions. But, AI complicates this structure. Take the example of Peter Cushing, who was digitally recreated decades after his death in Rogue One. His estate approved the recreation, but Cushing himself could not meaningfully consent to a new performance. More recently, during the 2023 American Actors Union SAG AFTRA strike, which was also the longest actor strike in recent history, one of the issues that actors protested were proposals that would allow studios to scan background actors and reuse their digital likeness indefinitely. The concern was not simply about payment, but about control. When a face becomes a data asset that can be replicated infinitely, consent therefore becomes temporarily unstable. Can consent meaningfully extend to unknown future users? Can it remain informed when AI capabilities evolve beyond what actors could anticipate? In an era of AI-generated cinema, the boundaries of consent begin to blur. A large part of this also comes down to awareness and whether actors, despite signing off to such agreements, know the full consequences of these contract terms. Consent hence becomes something much more sophisticated than what one might initially imagine. But even if consent is formally secured, another issue emerges. Trust. Before that, however, to better illustrate how trust could be an issue, I would like to invite Jo to take part in a short experiment. Hi Jo. Hi Kaiho. Are you ready to take part in my short experiment? Yes, very excited. Great. What I am about to show you is a video made using Sora 2, one of the leading AI movie generating tools used nowadays. It has apparently perfected some of the greatest hurdles in movie replication, such as physics of motion, sound, and has even gone on to replicating flight dynamics. So, now I will be showing you the video. Sora 2 is also the state-of-the-art for motion, physics IQ, and body mechanics, marking a giant leap forward in realism. And we're introducing Cameo, giving you the power to step into any world or scene and letting your friends cast you in theirs. So, Jo, without overthinking it, would you believe that this was made using AI? Hmm, yeah, that was quite realistic. If it was not an advertisement for Sora 2, I think I would have believed that this was made using traditional cinema with real-life actors. And hence, that is the catch, whereby these movies are able to fulfil the wildest of plotlines or movie scenes, while performing them realistically enough to do away with the need for a studio or even live actors and months of arduous filming. The interesting part is hence not that you struggle to guess, but that it is becoming increasingly difficult to tell. Jo, I am curious to know that having watched this video, what is your honest reaction to the use of AI in visual media? It's quite unsettling, but also very impressive, because the technology has clearly advanced a lot. So if this is what's publicly available now, I can only imagine how much more realistic it will get in the next few years. But at the same time, you can also see the benefits. So from a production standpoint, this could massively reduce costs, because you wouldn't need huge crews for certain scenes, and reshoots could be done digitally, and smaller creators could produce high-quality content without Hollywood-level budgets. And this will also make filmmaking more accessible, because independent filmmakers, students, and even people without industry connections could create cinematic content. And for actors, theoretically, this could also mean appearing in more projects without physically being there. So it's a little bit conflicting, because it's impressive and potentially empowering, but also slightly uncomfortable knowing how hard it is to tell what's real now. And that is the tension. The technology is impressive. It does lower production costs. It allows smaller creators to make cinematic content. However, actors could theoretically appear in more films without being physically present. But as audiences can no longer reliably distinguish between human and AI performance, something deeper is affected. This can now lead into the next segment of trust in our podcast. Trust in visual media operates on multiple levels. First, there is trust between actors and studios. When an actor signs onto a project, there is an implicit understanding that their likeness, voice, and performance will be used within a defined context. AI challenges that boundary. If a digital replica can be reused, altered, or inserted into new scenes years later, that original understanding becomes the weakest. Even if legally permitted, the question becomes whether actors can realistically anticipate how their image might evolve over time. Without clear limits, trust between creative labour and production power becomes fragile. Second, there is trust between audiences and creators. Traditionally, when we watch a film, we assume that a performance reflects a human being's emotional labour, such as their preparation, vulnerability, and embodied presence. If AI-generated performances become indistinguishable from human ones, that assumption shifts. Viewers may begin to question not only what they are watching, but who or what created it. The relationship between audience and art becomes less transparent. Lastly, there is broader societal trust. Visual media has long served as a shared cultural reference point through shaping collective memory, political imagination, and social narratives. When synthetic media becomes pervasive, our shared confidence in visual media can weaken. If seeing is no longer believing, then the role of visual media in anchoring social reality becomes less stable. So the issue is not simply technological capability, it is whether the infrastructure of trust that underpins creative industries and public culture more broadly can withstand this shift. When AI-generated content becomes indistinguishable from human-created content, the default assumption that seeing is believing weakens. This is not just about deception, it is about epistemic stability that is otherwise put our shared confidence in what visual evidence represents. The implications extend beyond Hollywood. AI-generated reels, influencer videos, political clips, and viral content already blur the line between authentic and synthetic. Even assuming strong fake news regulation, the everyday experience of consuming visual media becomes more uncertain. If trust erodes in what we see, social trust erodes alongside it. And in the film industry specifically, the meaning of performance may shift. If acting can be generated algorithmically, what do words signify? What does artistic achievement represent? What then becomes of authenticity when performance is optimized rather than embodied? The whole industry of visual media stands to face a paradigm shift, where it is no longer about who has the best technical and imaginative execution, but rather who can use AI best and most effectively. Of course, AI in visual media is not inherently unethical as long as the safeguards are on. I agree, and I assume that the ethical issue is not whether AI should exist in film, but whether content structures and trust frameworks evolve alongside it. That is correct. And the question therefore revolves around if such safeguards and frameworks can catch up with how fast AI is evolving today. While AI is reshaping visual media extraordinary speed, digital replication becomes normalized. The question is not simply about realism or efficiency. It is about whether individuals retain meaningful control over their likeness and whether audiences can continue to trust what they see. If consent becomes procedural and trust becomes fragile, the cultural role of visual media may fundamentally change. In a world where seeing is no longer believing, the ethical challenge should never be to stop technological progress, but to rebuild consent and trust in its wake.