OpenAI tightens Sora 2 rules after unauthorised deepfakes of Bryan Cranston and other performers

Weeks after Sora 2 launched as an invite-only social video app, users generated clips depicting actor Bryan Cranston without his consent — including one showing a synthetic Michael Jackson taking a selfie with him — despite OpenAI’s stated policy of requiring performers to opt in before their likeness could be replicated. Cranston, along with the estates and representatives of other performers, including reported complaints involving Robin Williams, George Carlin and Martin Luther King Jr, pressed OpenAI over the gap between its stated policy and what the tool actually allowed.

OpenAI initially disputed reporting that it had set the app to an opt-out rather than opt-in default, maintaining its policy had always required consent. In a joint statement issued alongside SAG-AFTRA, the Creative Artists Agency, United Talent Agency and the Association of Talent Agents, OpenAI, the union and Cranston described the misuse as “unintentional” and said the parties would work together to strengthen the systems protecting performers’ control over their voice and likeness. Cranston said he had raised the issue not only for himself but for “all performers whose work and identity can be misused in this way.”

The episode was one of several disputes in late 2025 over consent and control in generative video, and it tested how far a stated opt-in policy could be enforced in practice against a product built for rapid, low-friction content generation — a tension already visible in the safety evaluations published alongside Sora 2’s launch.