Meta Pulled Its AI Likeness Feature in Four Days. Here Is What That Actually Fixed and What It Did Not.
On July 7, Meta announced Muse Image, its first AI image generation model, with a rollout of more than thirty new effects across Instagram. Buried in that rollout was a feature that deserves to be studied in every conversation about likeness rights for years to come. It let any user @-mention a public Instagram account and generate AI images using that account as a visual reference. Your photos, your face, pulled into someone else's AI creation. Four days later, the feature was gone. Meta said it missed the mark. SAG-AFTRA called it a win. It was. But the speed of the reversal should not distract from what the launch revealed, because the thinking that produced this feature is still shaping products across the industry right now.
What the feature actually did
If your Instagram account was public and you were over eighteen, you were included by default. Anyone could reference your account and generate images of you. Meta excluded private accounts and minors automatically, which tells you the company understood the sensitivity. For everyone else, protection required finding a toggle in the sharing and reuse settings, and Meta's own documentation confirmed you would not be notified when AI content was created using your material. That structure has a name. It is opt-out consent, and calling it consent at all is generous. Genuine consent is informed, specific, and given before the use happens. A default you must discover and disable is enrollment, not agreement.
Why the industry moved so fast
CAA raised concerns directly with Meta and publicly called for protection to be the default, with individuals opting in if they want their likeness available for AI creation. SAG-AFTRA went further, publishing step-by-step opt-out instructions and urging not just members but all Instagram users to act. Within days, Meta disabled the feature. The pressure worked because the entertainment industry has spent three years building the vocabulary for exactly this fight. The 2023 SAG-AFTRA negotiations put digital replicas at the center of a historic strike. California passed AB 2602. The backlash to OpenAI's Sora earlier this year, which ran on a similar opt-out model before being shut down, gave everyone a fresh precedent. When Meta launched an opt-out likeness tool into that environment, the response was immediate because the arguments were already written.
The legal principle underneath all of it
Here is the through line I want you to take from this story. Consent to use a person's likeness is its own independent legal requirement. It does not travel with a platform's terms of service. It does not follow from the fact that a photo is publicly visible. And it cannot be manufactured by a settings default. This is the same principle at the heart of AB 2602, which took effect in January 2025. The statute requires that a performer give informed consent, with the terms clearly stated, before a digital replica can be used, and in key circumstances it requires the performer to have legal or union representation in the negotiation. The California legislature understood that consent extracted through boilerplate or buried defaults is not consent, and it wrote that understanding into law. A public Instagram profile is a decision to let people see your photos. It is not a license for anyone, including the platform itself, to turn your face into raw material for new synthetic content. Visibility and permission are different things, and the entire likeness rights framework depends on keeping them separate.
What the reversal did not fix
Meta withdrew one feature. It did not adopt opt-in consent as a design principle. The broader Muse Image rollout remains live, Muse Video has already been previewed, and every major platform is building comparable tools. The next version of this fight is a matter of when. And the reversal only happened because organized pressure made it happen. SAG-AFTRA could mobilize members within days. CAA could get Meta on the phone. If you are an independent creator, an influencer, or a performer working outside the union system, that infrastructure does not exist for you. Your protection comes from two sources: the law, which is real but incomplete, and your contracts, which are entirely within your control.
What to do now
First, check your own settings. Instagram's sharing and reuse controls govern whether your posts and reels can be used with Meta's AI features. Know where yours are set. Second, read the AI provisions in every agreement in front of you. Brand deals, collaboration agreements, production contracts, platform partnerships. If an agreement is silent on AI use of your likeness and voice, that silence is a gap, and gaps get filled by whoever holds more power. Third, if you perform in California, understand what AB 2602 requires before anyone can obtain rights to your digital replica. The law puts the burden on the party seeking your likeness. Hold them to it. The Meta reversal was a good week for likeness rights. But a controversy you have to win every time is not protection. Consent architecture written into law and into your contracts is. That is the work that continues after the headlines move on. Entertainment Lawyers of Los Angeles handles talent representation, production legal affairs, and creator rights, with a practice focused on AI and likeness protection. Free consultations available.

