OpenAI is limiting deepfakes on Sora after pressure from Bryan Cranston and Hollywood groups (www.businessinsider.com)

OpenAI said it’s working with actor Bryan Cranston, SAG-AFTRA and talent agencies to tighten limits on deepfakes in its Sora 2 video generator after Cranston’s voice and likeness were replicated without his consent. The company announced strengthened guardrails and an explicit opt-in policy for using anyone’s name, voice or likeness, expressed regret for “unintentional generations,” and committed to quick responses to complaints. The move follows earlier Sora changes to give more control to IP holders and a pause on AI-generated videos of Martin Luther King Jr. after the estate objected. For the AI/ML community this is a notable policy and product-level precedent: platform providers are implementing consent mechanisms and rights-management controls alongside technical mitigation. Key implications include the need for identity verification, provenance/watermarking and content-detection tooling, and closer alignment between model access, dataset curation and legal frameworks such as the NO FAKES Act. Developers of generative video and voice models should expect stronger opt-in/opt-out flows, enforcement pipelines for takedown requests, and ongoing coordination with rights holders as norms and regulations around synthetic media solidify.