Meta launched an AI feature that used your public photos — then killed it three days later

Information Technology 4 min 78 sources

A fast public "no" from an actor and an actors' union forced Meta to pull an image tool that opted everyone in by default. Plus: brain tech without the drill, glasses built not to watch, and a court that restored a hacked gamer's library.

Key takeaways

Meta launched an AI image tool that used public Instagram photos and was switched on by default, then pulled it within three days after an actor and the SAG-AFTRA union pushed people to opt out.
The week's other wins shared a theme of restraint and ownership: wearable brain tech that skips implants, smart glasses built with no camera, and a Brazilian court that made Microsoft restore a hacked gamer's game library.
The common lever across all of them is the default — who decides the starting terms of your relationship with a device or platform, and whether you were ever really asked.

Most weeks in tech, the momentum runs one way: a feature ships, the default is set to “on,” and the people it touches find out later. This week ran the other way. A few stories broke in the user’s favour — and each one turned on the same quiet lever: who gets to decide the terms.

Meta shipped it, then unshipped it

On Tuesday, Meta launched Muse Image — its first image-generation model from a group it calls Meta Superintelligence Labs — built into the Meta AI chatbot [5]. The feature let people generate and edit images using content pulled from public Instagram accounts [5]. The catch: it was switched on automatically for users, an opt-out rather than an opt-in [5].

The backlash was quick. Hannah Einbinder, the Emmy-winning actor from Hacks, posted on Instagram that the feature had turned itself on, and told people to go switch it off [5]. On Thursday, SAG-AFTRA — the union that represents actors and other media professionals — urged its members and Instagram users generally to opt out [5]. Its line was blunt: anything short of “a clear and conspicuous opt-in” for this kind of use was, the union said, “an utter miscalculation of public sentiment” [5].

By the end of the week, Muse Image was gone. “We’ve heard the feedback that this feature missed the mark, so it’s no longer available,” Meta said [5]. The company framed its original intent as giving people control over whether their public content could be referenced [5]. But the objection was never really about the AI. It was about the switch being on before anyone was asked. A feature that took three days to reach millions took three days to disappear — because a named voice and an organised one made the default the story.

The angle for anyone building products: the gap between “technically public” and “consented to reuse” is now a live liability, not a footnote. If your feature assumes yes, budget for the reversal.

Brain tech is trying to skip the skull

While the loudest brain-computer story stays Elon Musk’s Neuralink — which implants a chip inside the head — a quieter bet is that the mass-market version won’t require opening the skull at all [4]. A brain-computer interface, or BCI, reads electrical signals from the brain and turns them into commands, so a device can be controlled by thought [4]. CNBC profiled BrainCo, a Hangzhou startup building wearable versions that sit outside the head rather than inside it [4].

The therapeutic wins are already real: BCIs have let people with degenerative conditions such as ALS type and play video games using brain signals alone [4]. And the less-invasive path is gaining official ground — Chinese regulators recently approved what officials call the world’s first minimally invasive BCI cleared for commercial use, made by Neuracle Medical Technology, to help restore some hand function after spinal-cord injury [4]. Funding for the whole field is still a fraction of what flows into AI [4]. But the direction matters: a technology that once meant brain surgery is inching toward something you can put on and take off.

Two products that chose not to watch you

Two smaller stories carried the same restraint. Even Realities, whose G2 smart glasses show text on a green heads-up display, ships them with no camera and no speakers — on purpose [31]. The company’s pitch is that the glasses are for productivity, not recording, so the people standing in front of you don’t have to wonder if they’re being filmed [31]. It’s a deliberate refusal of the feature every other smart-glasses maker is racing to add.

And in Brazil, a court ruled against Microsoft after it told a gamer whose Xbox account had been hacked to simply re-buy his games [24]. The court ordered Microsoft to restore the account and its full library, and to pay about $400 in damages [24]. The ruling lands on a question most people never think about until it bites: when you “buy” a digital game, do you own it — or are you renting access the seller can revoke? A court, at least here, said the library was his to keep.

The small, unglamorous win: fixing your own stuff

To close on the least flashy story of the week: The Verge’s gadget newsletter spent its edition on a repair kit — the small tools that let you open, fix, and keep the things you already own instead of replacing them [33]. It’s easy to skip past. But the through-line of the week runs right through it. Whether it’s your photos, your face, your game library, or the phone in a drawer, the same question keeps surfacing: who decides what you can do with the thing that’s yours.

02 · Lesson · why it matters

The switch was already on

Whoever decides where a setting starts decides what most people will end up agreeing to — because almost no one goes looking for the toggle.

The fight was never about the machine

Meta built an AI tool that could make images from public Instagram photos. People were angry — but not really at the AI. Image generators are everywhere now. What set off an actor and a national union in three days was one small design choice: the feature was on before anyone was asked. It defaulted to yes.

That distinction is the whole lesson. The technology and the terms of the technology are two different things. You can be fine with the first and furious at the second. Meta lost the argument it didn’t know it was having — the one about the starting position of a switch.

The default is a decision that hides as a fact

Every feature ships in some state. Someone chose that state. On or off, opt-in or opt-out, visible or buried three menus deep — these are decisions, made by people, that then present themselves to you as just how the thing works.

And defaults are quietly powerful, because almost nobody changes them. This is one of the most reliable findings in how humans use technology: the option that comes pre-selected is the option most people keep. Not because they agree — because they never looked. When a company sets a feature to “on,” it is not predicting that most people will want it. It is arranging for most people to have it, whether they wanted it or not.

So “public content, used by default” isn’t a neutral description of how Instagram works. It’s a choice that serves the one who made it, dressed up as the natural order of things. The word “public” did a lot of hidden work there — it made “already yours to see” quietly slide into “mine to feed to a machine.” Those are not the same, and the gap is exactly where the anger lived.

The number that looked like success

Here’s the trap for the company. An opt-out default produces a huge number very fast. Overnight, millions of accounts are “enrolled.” On a dashboard, that looks like adoption — like people love it.

But it’s a hollow number. Most of those millions never chose. They were counted, not consulted. And a metric built on people who didn’t notice is fragile, because the moment they do notice, the same scale that made it look like a triumph makes it look like a violation. The bigger the auto-enrolled crowd, the bigger the reversal. Meta’s three-day round trip — millions in, millions out — is what happens when a success number turns out to be a consent number in disguise.

You are already inside this

It’s easy to file this under “a Hollywood spat.” It isn’t. If you have a public account anywhere — photos, posts, a face in someone else’s tagged picture — you are one of the millions the default was set for. The named actor got a headline. You got enrolled the same way, without the headline.

And it runs past photos. It’s the “buy” button that turns out to mean “rent,” until a court in Brazil says otherwise and hands a gamer back his library. It’s the smart glasses that ship with a camera pointed at strangers, until one maker ships a pair without. Each is the same question wearing different clothes: what did the starting position of this thing quietly decide on your behalf?

What the reversal actually shows

The hopeful part isn’t that Meta is good or bad. It’s that the terms turned out to be movable. A default that posed as fixed — this is just how it works now — got un-set in seventy-two hours, because enough people said the quiet part out loud at the same time.

But hold even that loosely. The people who set the default didn’t fully control the outcome either; public sentiment overran them. No single seat sees the whole board — not the user who never found the toggle, not the union that flipped it, not the company that thought a big number meant a happy one. The lesson isn’t that you can win every one of these. It’s smaller and more useful: the next time something arrives already switched on, that “on” was somebody’s decision, not the weather. Worth knowing where the switch is.

03 · Lab · your turn

Ship the feature

Rehearse how the default switch you set decides how many people get counted as "adopters" without ever choosing.

04 · Hope · carry this

The terms a giant sets are not the weather — this week a single clear voice and a few thousand people saying no together moved one in three days, which means the switches that arrive already flipped can still be flipped back.