Please Stop Making Me Opt Out of AI
WIRED Reece Rogers
Meta turned on an AI feature that let people tag Instagram accounts and generate images by default, so users had to opt out. After three days of public outcry, Meta pulled it — a rare speed record for a generative AI rollback.
Based on reporting by WIRED, Reece Rogers — read the original for the full story.
Summary, retelling and take written by AI under human oversight; images are AI-generated illustrations. How we work · Report an error
Early July brought a new Meta AI app feature that let anyone tag public Instagram accounts and generate images using those people's likenesses. Meta didn't ask users to turn it on. It just switched it on for everyone, leaving people to go hunting through settings if they wanted out. That choice, opt-out instead of opt-in, is what set off the trouble.
Instagram creators didn't wait around. Videos explaining how to disable the tagging feature spread fast, and the tone was frustration more than curiosity. Creator Sam Sooin Yang posted one such video that pulled in over 3 million views, saying flatly that she was tired of AI features being pushed on users who never asked for them. Three days after launch, Meta backed off, calling the feature a miss and rolling it back for its Instagram-linked AI chatbot.
Thorin Klosowski, a senior privacy activist at the Electronic Frontier Foundation, called the reaction unusually fast and said it was genuinely satisfying to watch play out that quickly. And it's worth sitting with that timeline: three days from a feature going live to a company quietly killing it. Few consumer tech backlashes move at that pace.
But Meta isn't the only company treating defaults as a growth lever. The author describes turning off Google's "Ask Gemini" bar the moment it appeared uninvited at the bottom of Google Docs, and says the same ritual has played out on Dropbox and LinkedIn. Ben Winters of the Consumer Federation of America argues Meta is just one particularly visible example of a broader pattern, one that persists partly because there's no strong federal privacy law forcing companies to do otherwise. He also points to a less flashy Meta setting worth checking: Facebook's Enhanced Browsing feature, which logs websites visited in-app on mobile.
Meta's response, delivered through spokesperson Daniel Roberts, leans on the idea that the company builds controls so people can shape their own experience, and that it funds research through groups like TTC Labs to make those controls easier to use. Woodrow Hartzog, who teaches law at Boston University, isn't buying defaults as neutral design. People tend to stay wherever they're dropped, he says, which is exactly why the starting position matters so much. He points to Article 25 of the EU's GDPR, which requires systems to collect only what's necessary and to pre-select the more privacy-protective option whenever one exists. Whatever its flaws in practice, that framework flips the entire logic Silicon Valley currently runs on.
My take — AI-written commentary, not fact-checked reporting
Nobody accidentally ships a face-tagging AI feature turned on by default — that's a deliberate bet that most people won't notice or won't bother fixing it before the goodwill damage is done. The fact that Meta killed it in three days doesn't prove the system works; it proves loud, coordinated backlash is now the only real check left, since nothing in US law forces the privacy-friendly option to come first. GDPR's default-to-privacy rule looks less like European red tape and more like the bare minimum a functioning market should require.
Read more about this at: WIRED