OpenAI says it is strengthening controls on its Sora 2 video generator after Bryan Cranston raised concerns about unauthorized replication of his likeness. The Breaking Bad star alerted SAG-AFTRA to the issue, which led to a joint effort between the union, OpenAI, and talent agencies to protect performer identities.
According to NBC News, OpenAI launched Sora 2 on Sept. 30 with rules against replicating real people without consent. Users needed explicit opt-in through a cameo feature. But videos of Cranston quickly appeared on the app alongside AI clips of other celebrities, including Michael Jackson and copyrighted characters like Ronald McDonald.
Union Collaboration Leads to Policy Changes
SAG-AFTRA, which represents more than 150,000 film and TV performers, worked with OpenAI and several talent agencies to address the problem. The groups issued a joint statement on Monday saying OpenAI has strengthened its original protections. CEO Sam Altman said the company is deeply committed to protecting performers from misuse of their voice and likeness.
Cranston said he brought his concerns to SAG-AFTRA after feeling deeply concerned for all performers whose work and identity can be misused. SAG-AFTRA President Sean Astin praised the resolution and said Cranston did the right thing by communicating with his union and professional representatives.
Copyrighted Content Also Blocked
The talent agency CAA had criticized OpenAI for exposing clients and their intellectual property to significant risk. Videos featuring copyrighted characters like Spongebob Squarepants, Pikachu, and Mario flooded the internet in the first weeks after launch. The Wall Street Journal reported that Sora 2 would let users generate copyrighted material unless copyright holders opted out.
New Guardrails Now Active
Requests to generate such clips on the Sora app now return an error message. The message states that the prompt may violate guardrails concerning third-party likeness or similarity to third-party content. OpenAI did not immediately respond to questions about whether it has changed its policy around copyrighted content.
The joint statement also expresses support for the NO FAKES Act, which aims to hold individuals, companies, and platforms liable for producing or hosting unauthorized deepfakes. The bill was introduced in the Senate in April but has not moved forward in Congress. Altman wrote that OpenAI was an early supporter of the legislation and will always stand behind the rights of performers.
Cranston said he is grateful to OpenAI for its policy and for improving its guardrails. He hopes all companies involved in this work respect the personal and professional right to manage replication of voice and likeness.
