OpenAI tightens Sora safeguards after celebrity deepfake backlash

openai

OpenAI has strengthened its guardrails around Sora, its AI-powered video generation app, following growing concerns over the unauthorised use of celebrity likenesses and voices. The company has now blocked users from creating videos that feature the faces or voices of real-life celebrities, including actors and public figures, without explicit consent.

The move comes after a joint statement was released by OpenAI, SAG-AFTRA, actor Bryan Cranston, United Talent Agency, Creative Artists Agency, and the Association of Talent Agents. The statement described a “productive collaboration” to ensure stronger protection of voice and likeness rights in AI-generated content produced through Sora 2 and the Sora app.

Actor Bryan Cranston had previously called out OpenAI after users were able to generate deepfake videos replicating his image and voice without his consent or compensation. The issue also drew criticism from the families of Robin Williams, George Carlin, and Martin Luther King Jr., who accused the company of allowing Sora to misuse the likenesses of iconic personalities.

While OpenAI already had an “opt-in” policy that required consent before simulating the likeness of living individuals, several users managed to bypass the system to recreate Cranston’s likeness. In response, OpenAI has tightened its safeguards to prevent any replication of voice or image without verified authorisation.

According to OpenAI, creators and individuals should have full control over how their identity and likeness are represented within AI systems. The company has also committed to responding swiftly to any future complaints or misuse reports.

This is not the first time OpenAI has had to act on concerns related to Sora. Late last week, it addressed similar complaints from the family of Martin Luther King Jr., promising stricter rules for depictions of historical figures. While OpenAI acknowledged the “strong free speech interests” in portraying deceased public personalities, it confirmed that estates or authorised representatives can now request the removal of such likenesses from Sora cameos.

Launched on September 30, Sora has quickly become one of the most downloaded AI apps on the App Store. However, this recent controversy underscores the growing tension between creative freedom and ethical responsibility in generative AI — a space where OpenAI is now taking a firmer stance.