Meta Just Made Your Face Raw Material for Anyone’s AI — and Opted You In by Default.

Most of our reporting this month has been about AI at a distance — chips, clouds, export controls, billion-dollar bets. Today's story is closer to home. It is about your face, your photos, and a default setting you were never asked about. And it is the clearest collision yet between the American way of building AI and the European way of governing it.

What Meta Launched

On July 9, Meta launched Muse Image, a feature built into its Meta AI assistant. The mechanic is simple and that is exactly the problem: a user types a prompt, references a public Instagram account with an @-tag, and receives an AI-generated image incorporating that account's likeness and visual content. As Wired described it, all someone has to do is tag your account in a prompt — if it is public — and they can use Meta AI to generate an image using your likeness.

The consent model is the flashpoint. Meta opts public accounts in by default. It does not ask the tagged person, and it does not notify them. The tool comes from Meta Superintelligence Labs, the AI division consolidated in June, and is rolling out on Instagram first before extending to Facebook, Messenger and WhatsApp.

The Backlash — Immediate and Broad

The condemnation arrived within a day, and from serious quarters. On July 10, the actors' union SAG-AFTRA urged its members and all Instagram users to opt out to protect their likeness.

The talent agency CAA condemned the opt-out approach, stating that no one's name, image, likeness, voice or creative work should be used by any third party, including AI models, without clear, documented consent. The advocacy group Public Citizen was blunter still, calling it an egregious invasion of privacy, with its technology policy director J.B. Branch saying Meta had chosen the creepiest possible path, and that people should not wake up to discover their face has become raw material for someone else's AI experiment.

Observers immediately drew the parallel to OpenAI's short-lived Sora video model, which operated on a similar opt-out basis before being pulled. The pattern is familiar: launch broad, make it opt-out, absorb the backlash, adjust only if forced.

Meta's Defence — Stated Fairly

In fairness, Meta's position deserves to be stated, not caricatured. Meta says every Muse Image output carries an invisible watermark it calls Content Seal, marking the image as AI-generated. It restricts the feature to adults' public accounts. And it does provide an opt-out: in Instagram settings, under Sharing and Reuse, a user can switch off allowing their content to be reused with AI features, or make their account private.

But each defence has a matching gap that the critics have been quick to identify. The Content Seal watermark establishes provenance, not consent — it confirms Meta's AI made the image, but does nothing to stop someone using your face in the first place. The opt-out is a buried setting most users will never find. And crucially, neither step is retroactive: Meta does not delete images the tool has already generated, and never notifies you when someone uses your photos. The safeguard, as one analysis put it, protects the product more than the person it exposes.

Why This Is a European Story Above All

Here is where gafam.ai's lens matters most, because the transatlantic difference is not rhetorical — it is legal, and it is stark.

Meta's entire model rests on opt-out by default. European data protection law rests on the opposite principle. Under the GDPR, consent must be freely given, specific, informed and unambiguous — an affirmative act. A pre-ticked box or a buried toggle that a user must hunt down to switch off is, almost by definition, not valid GDPR consent. A system that uses a person's facial likeness to generate images unless they discover and disable a setting inverts exactly the burden European law places on the company, not the citizen. What may survive, at least temporarily, in the American market runs into a wall the moment it touches European users.

The timing sharpens it further. As gafam.ai reported on July 1, the EU AI Act's Article 50 transparency obligations — including the labelling of AI-generated and manipulated imagery — activate on August 2, just three weeks from now. Meta's Content Seal watermark may speak to that labelling duty. But labelling an image as AI-generated does not resolve the prior question European law asks first: was there consent to generate it at all? And beyond the GDPR and the AI Act, many European jurisdictions have long-established personality and image rights — Germany's Recht am eigenen Bild, France's droit à l'image — that protect a person's control over the use of their own likeness independently of data protection.

For a European Instagram user, then, Muse Image is not an abstract governance debate. It is a US company deciding, by default, that their face is available as raw material — at precisely the moment European law is arming itself to say otherwise.

The European Perspective