News Desk: Why Attribution Rules Block the Requested Format

Transcript

SPEAKER_1: So last time we landed on a clear conclusion: the research supports the basketball, but it does not support putting specific words in a real public figure's mouth. Now I want to go deeper on why that line exists at all. SPEAKER_2: Good place to push. Because it is not just a style preference—there are two distinct conflicts in the supplied research. One is factual invention: we have no transcript of Rogan analyzing this playoff run. The other is persona imitation: mimicking a living public figure's voice and style carries its own separate set of risks. SPEAKER_1: So those are actually two separate problems, not one. SPEAKER_2: Exactly. Reuters reported that OpenAI updated its ChatGPT usage rules specifically to restrict content that impersonates real people or fabricates quotations. That policy addresses both issues at once—invented content and persona simulation. SPEAKER_1: And the BBC tied this to a broader regulatory push, right? Not just one company's internal policy. SPEAKER_2: Right. The BBC reported that major generative AI models are now trained under guidelines prohibiting audio or text that could be mistaken for real recordings of public figures. That came after US and EU regulators raised concerns about political deepfakes ahead of elections. SPEAKER_1: Think of the 2024 robocall incident the BBC flagged—an AI-generated fake of a prominent US politician's voice misled voters. Policymakers cite that episode constantly as the reason stricter rules exist now. SPEAKER_2: That is the concrete case that moved the needle. And it is why AP News reported that regulators and civil-rights groups have warned AI-generated deepfake audio of celebrities or journalists can mislead the public—prompting major platforms to restrict tools that explicitly impersonate real figures. SPEAKER_1: So what is the legal exposure specifically? The Guardian covered that angle. SPEAKER_2: The Guardian noted that defamation and privacy laws in the US and Europe expose publishers to liability if AI tools fabricate detailed false statements about identifiable individuals. That is why companies have built safeguards against invented quotations attributed to named people. SPEAKER_1: And the New York Times added the entertainment union dimension—performers pushing back on voice cloning without consent. SPEAKER_2: Yes. The Times reported that entertainment and media unions pressed technology firms to guarantee AI systems will not simulate performers' voices or personas without consent. Several AI providers publicly stated they will not mimic named entertainers as a result. SPEAKER_1: So for our listener—someone who came in wanting a fictional podcast-style show built around a living public figure's persona—the key idea is that the request runs into at least three converging constraints: factual invention, persona imitation, and legal exposure. SPEAKER_2: Three constraints, yes. Reuters also noted that several US states, including California and New York, introduced or expanded laws targeting unauthorized digital replicas of individuals' likenesses and voices. That reinforces platform policies against generating synthetic performances by specific real people. SPEAKER_1: And the Washington Post flagged the trust angle—podcast audiences getting confused when AI mimics a known host. SPEAKER_2: The Post reported that podcast creators and news outlets have expressed concern that AI-generated shows mimicking well-known hosts could confuse audiences and undermine trust. Some platforms have moved to bar AI from using living people's names in fictional dialogues entirely. SPEAKER_1: The FTC warning is worth naming here too. AP News covered that. SPEAKER_2: AP News reported that the US Federal Trade Commission warned AI providers that misrepresenting synthetic media as authentic—or allowing deceptive impersonations—could be treated as an unfair or deceptive practice under federal law. That is a significant enforcement signal. SPEAKER_1: So the alternatives remain what we outlined before: generic speakers, a neutral host-analyst format, clearly labeled fiction. The Guardian also noted that major publishers now require human editorial review of any AI-assisted scripts specifically to catch fabricated attributions. SPEAKER_2: That is the practical path. The European Union's AI Act—moving into implementation through 2025 and 2026, as the BBC reported—obliges providers of general-purpose AI systems to implement safeguards against illegal content, including certain forms of defamation and impersonation. The regulatory environment is tightening from multiple directions at once. SPEAKER_1: The takeaway for anyone producing this kind of show is that these pressures—legal, regulatory, platform policy, and editorial standards—are converging on the same answer. SPEAKER_2: many AI systems are now configured to refuse requests requiring fictional news-style dialogue using the real names and personas of living public figures, especially in formats resembling interviews or podcasts. The format is the problem, not just the content.