Meta's Muse Image draws SAG-AFTRA, CAA, Public Citizen fire

Meta shipped Muse Image inside Instagram and WhatsApp this week and, according to Axios, the mechanic that has drawn fire is small and specific: tag a public Instagram account in a prompt, and Meta AI will generate an image using that person's likeness. No affirmative consent, no notification to the person being depicted, and the opt-out lives inside Instagram's Sharing and reuse settings under AI features. The people whose faces are the raw material are the ones expected to go find the switch.

The pushback is unusually well-coordinated for a product that is only days old. SAG-AFTRA is telling its members, and all Instagram users, to opt out. CAA, one of the biggest talent agencies in Hollywood, said in a statement 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," and is asking Meta to make protection the default and let people opt in if they want their likeness used. Public Citizen's J.B. Branch, the group's director of federal AI governance and technology policy, called it an "egregious invasion" of privacy and said "People should not wake up to discover their face has become raw material for someone else's AI experiment."

Meta's response, from Superintelligence Labs chief Alexandr Wang, is that the company is "definitely receiving a lot of the feedback and are being thoughtful about what the next steps for that product should be." The company also says Muse Image has built-in protections designed to prevent generation of violent, sexual, or defamatory imagery of real people. That is a real safeguard for the worst outputs. It does not answer the default-consent question, which is the one the union, the agency, and the policy group are all asking.

The honest caveat is that the reporting does not give you the numbers a strategist would want: how many public accounts got auto-enrolled, how many have flipped the setting off since launch, or what exactly happens to AI images already generated of someone who later opts out. Take the specifics as reported, not settled.

The forward-looking piece is that consent defaults are becoming a competitive surface. If a talent agency, a union, and a Washington policy shop can all point at the same setting on the same product in the same week, opt-in likeness handling stops being a niche compliance choice and starts being a marketing position. Whoever ships that first, credibly, gets to sell it.