“While there are strong free speech interests in depicting historical figures, OpenAI believes public figures and their families should ultimately have control over how their likeness is used,” OpenAI said in a post on X from its official newsroom account. “Authorized representatives or estate owners can request that their likeness not be used in Sora cameos.”

The restriction comes just a few weeks after OpenAI launched its social video platform, Sora, which allows users to create realistic AI-generated videos resembling historical figures, their friends, and users who elect to have their likeness recreated on the platform. The launch has stirred fervent public debate around the dangers of AI-generated videos, and how platforms should implement guardrails around the technology.

Dr. Bernice King, Dr. King’s daughter, posted on Instagram last week asking people to stop sending her AI videos resembling her father. She joined Robin Williams’ daughter, who also asked Sora users to stop generating AI videos of her father.

The Washington Post reported earlier this week that Sora users had created AI-generated videos of Dr. King making monkey noises and wrestling with another civil rights icon, Malcolm X. Scrolling through OpenAI’s Sora app, it’s easy to find crude videos resembling other historical figures, including artist Bob Ross, singer Whitney Houston, and former President John F. Kennedy.

The licensor of Dr. King’s estate did not immediately respond to TechCrunch’s request for comment.

Beyond how Sora represents humans, the launch has also raised a flurry of questions around how social media platforms should handle AI videos of copyrighted works. The Sora app is also full of videos depicting cartoons like SpongeBob, South Park, and Pokémon.

OpenAI has added other restrictions to Sora in weeks since its launch. Earlier in October, the company said it planned to give copyright holders more granular control over the types of AI videos that can be generated with their likeness. That may have been a response to Hollywood’s initial reaction to Sora, which was not great.

As OpenAI adds restrictions to Sora, the company seems to be taking a more hands-off approach to moderating content in ChatGPT. OpenAI announced this week that it would allow adult users to have “erotic” chats with ChatGPT in the coming months.

With Sora, it seems that OpenAI is grappling with the concerns that come along with AI video generation. Some OpenAI researchers publicly wrestled with questions about the company’s first AI-powered social media platform in the days after its launch, and how such a product fits into the nonprofit’s mission. OpenAI CEO Sam Altman said the company felt “trepidation” about Sora on launch day.

Nick Turley, the head of ChatGPT, told me earlier this month that the best way to teach the world about a new technology is putting it out in the world. He said that’s what the company learned with ChatGPT, and that’s what OpenAI is finding with Sora, too. It seems the company is learning something about how to distribute this technology, as well.