Meta's AI Chief Posted a Deepfake of Altman to Promote the Muse Agent
Meta AI chief Alexandr Wang posted a generated video on Instagram promoting Muse, Meta's AI agent for everyday users. In the clip, he, Zuckerberg, and OpenAI's Sam Altman hold a video call, and Altman is dressed as Muse's fluffy mascot. The post is labeled as AI, because Meta requires labels for videos with people, but not for images: a fake photo of Mia Goth spread without any label. Product teams should label all formats the same way.
AI-processed from The Verge; edited by Hamidun News
Imagine this: the AI chief of one of the largest tech companies posts a generated video in which his main competitor is dressed as a cartoon creature. A couple of years ago, a deepfake, meaning a fake video of a real person, would have prompted loud statements and threats of lawsuits. Now it's a marketing move. In November, Alexandr Wang, Meta's head of AI, did exactly this, and the story shows how fast the boundaries of what's acceptable have shifted and where social platforms still have gaps in their rules.
What the video Alexandr Wang posted contains
Wang published a short clip on Instagram to promote Muse. It is Meta's consumer AI agent, meaning an assistant for ordinary people that carries out tasks on the user's behalf. In the video, Wang, Meta CEO Mark Zuckerberg, OpenAI CEO Sam Altman, and Jolly, Muse's humanized mascot, talk over a video call, four of them in total. When Altman joins the call, he looks like Jolly: a fluffy hooded costume and makeup in the style of a cute character.
The caption reads: "When the bros discuss openai's product strategy." It's a joke about the race between AI agents and Muse's recent rise in popularity. Wang is promoting the product in what people call an "anti-PR" style: he has written hundreds of tweets and calls traditional marketers "corposlops." For a top executive at this level, the approach is unusual, and the post directly took a jab at a competitor.
Where the labeling line is drawn
Meta requires AI labels on videos that show people. Wang's post carries the label. Meta spokesperson Tracy Clayton pointed this out. He added that Altman himself opened access to his likeness for content generation. That was the case when OpenAI launched the Sora app, and it was quickly flooded with fake videos of Altman. OpenAI did not respond to requests for comment.
There is no such requirement for static images. This became evident in October with a generated image of actress Mia Goth. It gained popularity with no disclosure that a neural network had created it. The picture was distributed by a fan account with the blue verification checkmark that Instagram itself grants. A representative for the actress told The Verge that it is "incomprehensible" to him how the platform allowed the image to spread instead of caring about "the protection and integrity of a public figure's image."
Scarlett Johansson also regularly becomes the subject of sexualized AI images. Meta has previously removed some accounts that impersonated other people and used their appearance. But these were reactions to individual cases, not a general rule applied across all formats.
What this means for developers and product managers
First: AI content has become part of the ordinary environment, especially on social networks. Ahead of the US midterm elections, political groups spent millions of dollars on generated ads against their rivals. The reaction "this is unacceptable" has turned into "as long as it's labeled."
Second: labeling rules on the platform are uneven. Videos with people must be labeled; images need not be. As a result, the same fake gets different treatment depending on the format. If you are building a product with user-generated content or image generation, decide in advance whether you will label all formats equally. A different approach to video and photos creates a loophole, as in the Mia Goth case.
Third: a person's consent to the use of their likeness increasingly serves as an argument in a dispute. Meta points to the fact that Altman allowed his face to be used in Sora. But that permission applies to a specific service. Whether it covers a competitor's promotional video on another company's social network is not addressed in the report, and for product managers this is an area of legal risk.
The story has one limitation: it rests on a few public examples, not on platform-wide statistics. The scale of the problem on Instagram has not been measured here. But the signal for the industry is clear. When executives themselves publish AI videos featuring real people, pressure on platforms and generative model developers will grow. It is better to plan labeling, account verification, and a response to complaints in advance than to answer claims after the fact.
- Nov 21, 2026Open-Weight Inkling 2 Mini Matches Qwen3.8-27B on Independent Benchmark, Runs on 16 GB of Memory
- Nov 21, 2026Qwen3-0.6B Fine-Tuned on a Mac mini: Errors Dropped to 3 out of 30 from 9
- Nov 21, 2026Thinking Machines Releases Open-Weight Inkling 2 Mini with 26.4B Parameters for 32 GB Macs
- Nov 21, 2026Meta Is Reportedly Preparing the $300 Muse Charm AI Tamagotchi With a Year of Verizon 5G
Key Questions Answered
What the video Alexandr Wang posted contains?
Wang published a short clip on Instagram to promote Muse. It is Meta's consumer AI agent, meaning an assistant for ordinary people that carries out tasks on the user's behalf. In the video, Wang, Meta CEO Mark Zuckerberg, OpenAI CEO Sam Altman, and Jolly, Muse's humanized mascot, talk over a video call, four of them in total. When Altman joins the call, he looks like Jolly: a fluffy hooded costume and makeup in the style of a cute character.
Where the labeling line is drawn?
Meta requires AI labels on videos that show people. Wang's post carries the label. Meta spokesperson Tracy Clayton pointed this out. He added that Altman himself opened access to his likeness for content generation. That was the case when OpenAI launched the Sora app, and it was quickly flooded with fake videos of Altman. OpenAI did not respond to requests for comment. There is no such requirement for static images. This became evident in October with a generated image of actress Mia Goth.
What this means for developers and product managers?
First: AI content has become part of the ordinary environment, especially on social networks. Ahead of the US midterm elections, political groups spent millions of dollars on generated ads against their rivals. The reaction "this is unacceptable" has turned into "as long as it's labeled." Second: labeling rules on the platform are uneven. Videos with people must be labeled; images need not be. As a result, the same fake gets different treatment depending on the format.
Test Frontier LLMs & Autonomous Agents in One Workspace
Run Claude Opus 5.5, GPT-6, Gemini 3.1, and DeepSeek side-by-side in Nomad AI (nomadnet.ai) — unified model routing, deep research, and multimodal workflows under one subscription.
The AI world, distilled — once a week
Seven stories that actually mattered, hand-picked. No noise, no reposts, no press releases.
Done! Check your inbox for a confirmation.