Meta’s Muse Image echoes the same privacy fights that hit Grok and Sora

Meta is arriving late to a party that hasn’t gone well for anyone. Elon Musk’s X wired image generation into its social platform two years back, and last holiday season it turned into a meme to get Grok to spit out sexualised images of people, including minors. OpenAI’s Sora required users to opt in before their likeness could appear in a video — a safeguard Meta’s new Muse Image tool reverses entirely — and Sora shut down in March anyway, partly over cost.

Muse Image, launched Tuesday out of Meta Superintelligence Labs, lets anyone tag a public Instagram handle and generate AI images from that account’s published photos and reels, with every adult public account automatically opted in and no notification sent when someone remixes their face.

Pressed on safeguards, a Meta spokesperson said the tool blocks violent, sexual or defamatory imagery of real people, and that content breaking Community Standards gets enforced. Underage public accounts are walled off from the feature, and teens cannot use it on others.

Meta AI chief Alexandr Wang framed the launch as a creativity unlock — custom event invites, mock-ups and personalised graphics ready to post — but the tool’s capability, which sits at No. 2 on the Arena leaderboard for text-to-image, isn’t the concern. It’s that the feature works on strangers by default, with the burden of opting out falling entirely on the user.