Skip to content
Theme
OnDevice Tools

Internet & SocialSocial Media

Instagram AI deepfake: Meta's AI chief posts fake Sam Altman clip

Meta requires AI videos of real people to be labelled, but the same does not apply to still images. That gap has already affected celebrities on the platform.

Illustration of a phone showing an AI video call with Sam Altman in a furry mascot costume

Key takeaways

  • Alexandr Wang, Meta's chief AI officer, posted an AI-generated clip of Sam Altman to promote Muse, according to The Verge.
  • Meta requires AI videos showing real people to be labelled as AI, but still images need no such label.
  • An AI image of Mia Goth spread in October from a verified fan account with no disclosure.
  • Meta says Altman has permitted use of his likeness in AI content; OpenAI did not comment.
In this article
  1. A four-way call with a costumed Altman
  2. Where Meta's AI labels stop
  3. Celebrities targeted by fake images
  4. How norms around AI imitation shifted
  5. What to check before sharing

A four-way call with a costumed Altman

Instagram AI deepfakes are no longer only the work of strangers and fan accounts. Earlier in November, Alexandr Wang, Meta's chief AI officer, posted an AI-generated clip showing OpenAI chief executive Sam Altman in a furry costume, according to The Verge, which reported the post on 17 November 2026.

The short clip was published to promote Muse, Meta's consumer AI agent. It shows a four-way video call between Wang, Meta chief executive Mark Zuckerberg, Altman and Jolly, the cartoon mascot of Muse.

When Altman joins the call, he is dressed as Jolly, wearing a furry hooded outfit and face paint to match the character. The caption joked about OpenAI's product strategy. The Verge reads it as a nod to the contest between AI agents, which are software tools built to carry out tasks for users, and to the recent attention on Muse.

Where Meta's AI labels stop

Meta's rules require AI-generated videos that depict people to be marked as AI, and Wang's post carries such a label. The requirement does not extend to still images, according to The Verge.

That gap mattered in October, when an AI-generated image of the actress Mia Goth spread widely on Instagram with no disclosure. The picture was shared by a fan account that held a blue verification badge from Instagram.

Goth's representative told The Verge that allowing the image to spread was "incomprehensible", and argued that the platform should instead focus on protecting a public figure's likeness.

Celebrities targeted by fake images

Goth is not the only public figure affected. The Verge reports that other celebrities, Scarlett Johansson among them, are regularly the subject of sexualised AI imagery.

Meta has previously removed accounts that it found were posing as other people and misusing their likeness. The company has not said whether it will act on Wang's post, which follows its own labelling rule.

The Altman clip is lighter in tone than the images described above, but it relies on the same technique: a synthetic likeness of a real person, made without a camera and shared with a large audience.

How norms around AI imitation shifted

Illustration: AI, OpenAI logo, Meta logo, Instagram logo and the labels AI Deepfakes, Alexandr Wang, Instagram

A few years ago, AI content imitating real people drew strong public criticism, and in some cases threats of legal action. Such material is now part of everyday social media, The Verge notes. Political groups ran AI-generated attack ads worth millions of dollars ahead of the midterm elections.

Wang has promoted Muse in a deliberately informal style that some observers have described as anti-PR. He has posted hundreds of times and used a mocking nickname for conventional marketers.

Meta spokesperson Tracy Clayton told The Verge that Wang's post is labelled as AI content. She also pointed out that Altman has allowed his likeness to be used in AI material, as happened when OpenAI launched its Sora app and the service was soon filled with fake Altman clips. OpenAI did not respond to requests for comment.

What to check before sharing

Because Meta labels AI video but not AI images, viewers can look for an AI marker before treating a clip as genuine footage. A convincing still image of a public figure may arrive with no such warning.

A blue badge on an account is not proof that the person shown created the post. In the Goth case, the badge belonged to a fan account.

Readers who suspect impersonation can report the account to Instagram. Meta says it removes accounts that it finds are posing as someone else and using their likeness.

Sources

  1. Not even Sam Altman is safe from AI deepfakes on InstagramThe Verge

Tags

Spotted a mistake? See our corrections policy. Read how News by OnDevice works in our editorial policy.