Meta Kills Its AI Deepfake Feature After Public Backlash

It lasted about as long as you would expect. Meta has pulled a feature from Instagram that allowed users to generate AI images of public figures, including celebrities and creators, after the backlash grew loud enough that the company apparently decided the heat was not worth it.

The feature, which was part of Instagram's broader AI image generation tools, let users create photorealistic images of real people by typing in prompts. Critics pointed out almost immediately that this was essentially a built-in deepfake machine, and that targeting public figures did not make it less harmful — it arguably made it worse, because those images would circulate with the implicit credibility of having been made inside an official Meta product.

What the Feature Actually Did

The tool drew on Meta's generative AI infrastructure to produce images based on text prompts. Users could, in theory, create convincing fake images of celebrities, politicians, journalists, or anyone else with a public profile. Meta had some guardrails in place, but guardrails around AI image generation have a poor track record of holding. Researchers and journalists have consistently demonstrated that content policy filters can be nudged, reworded around, or simply failed by the underlying model.

The specific concern here was not just that users could create fake images, but that they could do so easily, at scale, and through an app with over two billion users. The mainstream accessibility of deepfake creation has always been the core problem — when it requires technical skill and specialist software, it stays in a relatively small community. When it is three taps away on Instagram, the calculus changes entirely.

Public figures pushed back hard. Several creators and celebrities made clear they had not consented to being used as training or generation targets. The consent question is central here: there is no mechanism by which a public figure can opt out of being rendered in AI-generated scenarios, and Meta had not built one in.

Meta's Pattern With Controversial AI Features

This is not the first time Meta has launched an AI feature, faced significant criticism, and then quietly walked it back. The company has been aggressive in deploying generative AI across its platforms, and it has occasionally moved faster than its own trust and safety teams can keep up with. That is a product strategy decision, not an accident.

What makes this particular rollback interesting is the speed of it. Meta pulled the feature after backlash, not after a regulatory intervention, not after a lawsuit, and not after a prolonged public investigation. That suggests internal concern about where this was heading. The EU's AI Act and various national regulations around synthetic media are getting teeth, and Meta's legal team will be aware that AI-generated non-consensual imagery of real people is exactly the territory those regulations are designed to cover.

The UK, for example, now criminalises the creation of intimate deepfakes under the Online Safety Act. While the Meta feature was not specifically positioned for that use case, the infrastructure it provided could easily have been pointed in that direction. Regulators tend to look at what a tool enables, not just what it was marketed as.

What Happens Next

Meta has not said it is abandoning AI image generation entirely, which would be commercially absurd given how much it has invested in this space. The more likely outcome is that the feature returns in a modified form, with tighter restrictions on which individuals can be depicted and possibly with an opt-out registry for public figures.

The broader question is whether any guardrails are sufficient for a tool that, by design, creates photorealistic fake images of real people. The technology is not going away. But the gap between what these models can do and what platforms are currently willing to deploy is closing, and stories like this one are a big part of why.