OpenAI cracks down on deepfakes after actors’ union concerns

OpenAI said it will step up efforts to crack down on deepfakes on its video app Sora after concerns from actor Bryan Cranston and SAG-AFTRA, the US actors’ union.

Cranston raised the issue after unauthorized AI-generated videos using his voice and likeness appeared on Sora following its late September launch.

Cranston’s agency, United Talent Agency, and Creative Artists Agency have criticized Sora for risks to clients’ intellectual property.

OpenAI said it will work with SAG-AFTRA, UTA, the Association of Talent Agents, and Creative Artists Agency to strengthen protections against unapproved AI-generated content.

🔗 Source: CNBC

🧠 Food for thought

Implications, context, and why it matters.

OpenAI tightens Sora 2 guardrails, but tests find gaps

OpenAI pledged to “strengthen guardrails”, yet NewsGuard tests got Sora 2 to make realistic videos pushing false claims in 80% of tests and covering all five Russian disinformation narratives 1.
Free online tools removed Sora watermarks in four minutes 1. One test produced a clip resembling a public figure after an indirect description, though attempts were blocked 1.
After a request from King’s estate, the company blocked Martin Luther King Jr. videos on Sora. It pledged faster complaint response. Partners now include SAG-AFTRA (the Screen Actors Guild–American Federation of Television and Radio Artists) and United Talent Agency. The Association of Talent Agents and Creative Artists Agency are participating.

Platform liability opens a compliance market as NO FAKES Act advances

Backed by OpenAI, NO FAKES Act would set notice-and-staydown. After a rightsholder flags an infringing replica, the platform must stop reuploads and block matching uploads 2. That shift drives demand for compliance tech to fingerprint, hash, and match content.
Vendors can offer likeness-detection tools (to identify a person’s voice or face); Application Programming Interfaces (APIs); consent management platforms (systems that capture and track permissions); and automated content moderation tools. Customers include AI video platforms, social networks, and ad networks facing statutory damages up to $750,000 per unauthorized digital replica 2.
The bill covers many online services 2. Covered groups span app stores; search engines; ad networks; e‑commerce; mapping; hosting; and cloud providers that allow public access to user‑uploaded content.