Seventy-Two Hours

Meta launched the first product from its Superintelligence Labs on July 8, 2026. They called it "Muse Image." It was an AI-powered photo editor for Instagram that could apply 30 filters to any photo, including photos posted by other people. The feature was on by default. You could @mention a friend's public account and generate AI-edited images of them. Birthday cards, memes, "playful edits between friends," as Meta put it[1].

Seventy-two hours later, they pulled it.

The backlash was immediate and predictable to everyone except, apparently, the people who built it. SAG-AFTRA, the actors' union, condemned the product, saying that "anything other than a clear and conspicuous OPT-IN for these types of uses of Instagram users' images is unacceptable"[2]. Users raised the obvious question: what stops someone from creating damaging or humiliating fake images of you using your public photos? Meta's answer, effectively, was trust us. The internet's answer was no.

What makes this story remarkable is not the failure itself. Products get pulled all the time. What is remarkable is the gap between the marketing and the judgment. This came from Meta Superintelligence Labs, a division explicitly tasked with building "a personal superintelligence that knows us deeply, understands our goals, and can help us achieve them"[1]. That is a high bar. The first product out of that lab was a feature that let you slap a filter on someone else's selfie without asking.

Meta leads the world in three things: the number of people on its platforms, experience with people behaving horribly online, and dealing with community backlashes after privacy abuses. And yet nobody in the product chain apparently raised the flag that "edit anyone's public photos with AI, enabled by default" might be a problem[1].

There is a pattern here, and it is not unique to Meta. A company launches a feature that puts convenience and engagement ahead of consent. Users push back. The company retreats, issues a statement about "hearing the feedback," and frames the withdrawal as responsiveness rather than a lapse in judgment. We have seen this cycle with facial recognition, location tracking, ad targeting, and now AI-generated images of real people.

The fix is always the same and always simple: make it opt-in. Let people decide whether their face can be used in someone else's AI generation. The technology is not the problem. The default is the problem. When the default is "anyone can use your image," you are not building superintelligence. You are building a liability with a filter on top.

Meta said the feature was designed by Instagram creators who "used Meta AI to amplify their creativity and bring their ideas to life." That is nice. But the road to a 72-hour product recall is paved with creative ideas that skipped the part where you ask whether something should exist in the first place.

Superintelligence is a big word. It implies foresight, restraint, an understanding of consequences. None of those were on display here. The first product from a lab named after superintelligence was a photo filter that could not survive a weekend. That tells you something about how far we are from the thing they are naming their division after, and how close we still are to the old playbook of move fast, apologize later.