Meta’s Muse Image AI sparks Instagram privacy backlash

TL;DR:

• Meta’s new Muse Image tool lets users generate AI pictures using the faces of anyone with a public Instagram profile, by default and without notifying them.
• Privacy groups including the EFF, Proton and Malwarebytes say the feature should be opt-in, not a setting buried in menus.
• The tool is live in the US and limited markets, with Meta planning to extend it to Facebook and add video.

Meta is facing criticism from privacy advocates after its latest AI image generator, Muse Image, was found to pull faces from public Instagram profiles by default. Users can tag a public account in a prompt and generate pictures drawing on that person’s posts — and the person is never told.

Opt-out by design

Meta says private accounts and under-18s are excluded automatically, and that adults with public accounts can switch the feature off through a “sharing and reuse” toggle in settings. Critics counter that placing the burden on users is the problem. Proton warned that “data sharing is turned on by default, the opt-out is buried deep in settings, and public backlash becomes the main way users find out what happened to their content”, while Malwarebytes noted that simply finding the setting “is its own adventure”. The Electronic Frontier Foundation’s Thorin Klosowski argued the capability “should absolutely be opt-in”, calling it a new use of photos people posted long before such tools existed.

A specific worry is children. Meta says under-18s cannot be tagged, but has not clarified whether children pictured in adults’ public posts could be fed into prompts — a gap that echoes the UK’s ongoing review of safety rules for AI-enabled products. For UK Instagram users the tool is not yet widely available, but Meta’s default-on approach sets a template that regulators under the Online Safety Act and UK GDPR are likely to scrutinise.

Looking forward

Muse Image currently runs in the Meta AI app, Instagram Stories in the US and WhatsApp in limited countries, with Facebook and video features planned. The blunt fix for the worried is to make an account private; the more measured route is to hunt down the opt-out. Either way, the episode reopens a familiar question about consent — whether “public” posts from years ago can fairly become raw material for generative AI without the subject’s say-so.