Public sharing isn't consent to being remixed by machines
When Meta quietly activated its Muse Image feature across millions of Instagram accounts this week, it set off a confrontation that has been building for years — between the vast machinery of AI development and the quieter human assumption that sharing something publicly does not mean surrendering it to transformation. The company's swift reversal, following condemnation from legal scholars, journalists, and Hollywood's most powerful talent agency, reveals how fragile the social contract around digital consent remains. What lingers after the feature's removal is not resolution, but a sharper version of the original question: in the absence of law, what truly protects what we choose to share?
- Meta launched Muse Image with fanfare but buried the critical detail — it was switched on by default for millions of users, with no warning and no easy way out.
- Legal scholars, AI researchers, and media outlets moved quickly, arguing that public sharing has never meant consent to be remixed, fed into a machine, and returned as something unrecognizable.
- Hollywood's most powerful talent agency issued a pointed rebuke, making clear that even the world's biggest stars had not consented — and would not — to their images and likenesses being absorbed by AI systems.
- By Saturday, Meta and Zuckerberg conceded, pulling the feature and offering a muted explanation that critics noted still failed to acknowledge the core ethical failure of opt-out over opt-in.
- The episode closes without resolution — experts warn that without government regulation, the same logic that produced Muse Image will simply resurface in a different form.
Meta unveiled Muse Image earlier this week, a tool that would allow users to generate AI images trained on public Instagram posts. The company framed it as a creative offering built from the world its users had already shared. What it did not emphasize was that the feature activated automatically for millions of accounts. Only private accounts and those belonging to minors were exempt from the start — everyone else had to find their way through settings to opt out.
The response was swift and pointed. Legal and AI experts in New Zealand and beyond argued that choosing to share something publicly carries no implication of consent to have it remixed and fed into an AI system. Andrew Lensen of Te Herenga Waka Victoria University called the approach unethical, stressing that most users would never know it was happening — and called for government regulation to prevent it. Cassandra Mudgway of Canterbury University echoed the concern, noting that public images are already being harvested by third-party AI systems in ways users never anticipated.
Hollywood added its voice through the Creative Artists Agency, which represents figures including Tom Cruise and Zendaya. Their statement was unambiguous: no name, image, likeness, or creative work should be used by AI without explicit, documented consent. Meta had been planning to extend Muse Image across Facebook, Messenger, and WhatsApp — that expansion now hung in doubt.
By Saturday afternoon New Zealand time, Meta pulled the feature entirely, saying it had missed the mark. Critics noted the explanation still sidestepped the central issue: the company had offered control only to those who actively sought it out. The episode joins a growing list — including OpenAI's withdrawal of Sora after deepfake concerns — that points toward the same unresolved question: without regulation, what prevents the next attempt?
Meta announced a new feature called Muse Image earlier this week with considerable fanfare. The tool would let anyone generate images using artificial intelligence trained on public Instagram posts—photos, artwork, moments that people had chosen to share with the world. The company framed it as a creative gift to its users: image generation built for your world, they called it. There was one catch, though one Meta didn't emphasize in its rollout. The feature switched on automatically for millions of Instagram accounts. If you wanted to keep your public posts out of the system, you had to hunt through your settings and turn it off yourself. Only accounts marked private, and those belonging to people under 18, were spared from the start.
Within days, the backlash arrived with force. Journalists at outlets including the BBC, Wired, and the New York Times began asking hard questions. Legal scholars and AI researchers in New Zealand and beyond called the approach deeply troubling. Cassandra Mudgway, a senior law lecturer at Canterbury University who studies online abuse, put it plainly: choosing to share something publicly does not mean you have meaningfully consented to having it remixed, transformed, and fed into an AI system. Andrew Lensen, an AI specialist at Te Herenga Waka Victoria University, went further. He called the entire approach unethical. The real problem, he argued, wasn't that the feature existed—it was that most people would never know it was happening, or wouldn't have the time or energy to dig into their privacy settings to stop it. He called for government regulation to prevent this kind of behavior.
Hollywood weighed in too. The Creative Artists Agency, which represents major stars including Tom Cruise, Brad Pitt, and Zendaya, issued a statement making clear that no one's name, image, likeness, voice, or creative work should be used by any third party, including AI systems, without explicit, documented consent. The statement was a direct rebuke to Meta's approach. The company had been planning to roll out Muse Image across Facebook, Messenger, and WhatsApp as part of a broader push to weave AI into all its platforms. That expansion was now in question.
By Saturday afternoon, New Zealand time, Meta and CEO Mark Zuckerberg capitulated. The company posted a brief statement on its announcement page: the feature had missed the mark, they said. It was no longer available. In their explanation, Meta claimed their intent had been to provide a useful creative tool and to give people control over whether their public content could be referenced. The reality, however, was that they had given people control only if those people actively sought it out—a distinction that mattered enormously to the experts who had criticized the rollout.
The episode reflects a larger tension in how AI systems are trained and deployed. Public images are already being used by third-party AI systems in ways many users never anticipated or consented to, Mudgway noted. Tools like OpenAI's Sora had raised similar concerns by allowing dead celebrities and others to be used in deepfake generation, prompting OpenAI to pull the product from the market. Meta's quick reversal suggests the company understood the political and reputational cost of pushing forward. But it also leaves unresolved the fundamental question that Lensen raised: without government regulation, what stops companies from trying again?
Citações Notáveis
Just because someone has chosen to share content publicly does not mean they have meaningfully consented to it being remixed, transformed or reused by AI systems— Cassandra Mudgway, Canterbury University senior law lecturer
Having to opt-out is not ethically acceptable, especially when most people will be unaware or lack time to adjust their settings— Andrew Lensen, Te Herenga Waka Victoria University AI lecturer
A Conversa do Hearth Outra perspectiva sobre a história
Why did Meta think this would work? They must have known people would object to their images being used without asking.
They probably believed that public sharing was consent enough—that if you posted something to Instagram, you'd accepted the risks. But that's not how consent works in law or ethics. Sharing with your followers isn't the same as giving permission to a corporation to feed your image into a machine.
So the opt-out versus opt-in distinction is the real issue here?
Exactly. Opt-out means most people never change anything. They don't know, or they don't have time, or they forget. Opt-in would mean Meta had to ask first. The difference is enormous—it's the difference between presumed consent and actual consent.
Did Meta genuinely not see this coming, or did they hope no one would notice?
Hard to say. But the speed of the backlash—journalists, lawyers, Hollywood, AI experts all pushing back within days—suggests Meta underestimated how much this particular line would bother people. Or maybe they thought the convenience of the feature would outweigh the privacy concerns.
What happens now? Does this mean AI training on public data stops?
No. The broader problem Mudgway mentioned—that public images are already being used by AI systems in ways people didn't expect—that continues. This reversal is just one company stepping back from one feature. The real question is whether governments will step in and set rules about how AI can be trained on human-created content.