Your Image Is Yours. Meta's Muse Forgot the Case That Proved It.

An 1862 French court ruling and a 2004 Zuckerberg quote explain why the AI feature collapsed in days.

July 11, 2026

Last week Meta launched an AI feature called Muse Image. It let anyone generate images by tagging a public Instagram account, using that person's photos as raw material for AI "remixes," with no notice and no opt in required. Within days, after backlash from CAA, SAG-AFTRA, and users who found out their photos were fair game by default, Meta pulled the feature and admitted it "missed the mark."

I don't think Meta missed anything. I think we're watching a pattern that's over 160 years old on one side of the ledger and about two decades old on the other, and it's worth naming both halves plainly.

Muse Let Anyone
Remix Your Photos
Without Asking

Meta rolled out Muse Image through its dedicated AI unit, Meta Superintelligence Labs, earlier this month. One feature let users generate images by @-mentioning any public Instagram account, turning that account's photos into source material for AI remixes, unless the account holder had already found the setting and turned it off. Nothing alerted people that their photos were being used this way.

The backlash was immediate. CAA, whose clients include Tom Hanks and Meryl Streep, told Meta directly that no one's name, image, likeness, voice, or creative work should be used by a third party, AI included, without documented consent. SAG-AFTRA raised similar concerns. Within days, Meta pulled the @-mention functionality and posted a short update: the feature "missed the mark, so it's no longer available." CAA called the reversal swift and commendable.

It's worth noting what this walkback is, and isn't. Meta conceded a trust and design failure. Whether the original feature was also unlawful use of likeness is a separate question, one still being worked out through pending legislation, not one this incident alone settles. I'm not arguing the legal case here. I'm arguing the ethical one, and the ethical case doesn't need the legal one to be finished first.

If this shape looks familiar, it should. OpenAI's Sora app launched last year with thin IP protections, quickly filled with unauthorized celebrity likenesses, and was walked back within days under similar pressure before OpenAI shifted its strategy elsewhere entirely. Launch broad, absorb the backlash, retreat just far enough to lower the temperature. That's not a bug in how these companies operate. It's the operating model.

In 1862, France Ruled That
an Image Belongs to the Person Who Made It

In November 1862, France's highest court, the Cour de Cassation, ruled on a dispute between the society photography studio Mayer et Pierson and a pair of rivals, Betbéder and Schwalbé, who had been reproducing and selling the studio's portraits without permission, including a well known portrait of the Count of Cavour. A group of prominent painters, including Ingres, had signed a public declaration arguing that photography was mechanical reproduction, not art, and therefore didn't deserve copyright protection at all. The court disagreed. It ruled that a photograph could carry the personal imprint of its author and was entitled to the same protection as any other work of art.

That ruling did more than settle a dispute between two Parisian studios. It established, as a matter of legal principle in France and eventually far beyond it, that an image belongs to the person who made it. Not to whoever can get their hands on it and reproduce it more cheaply. Not to whoever finds a clever technical justification for why copying doesn't count this time. The photographer.

CAA's statement about Muse, that artists deserve to decide if and how their likeness and work is used, with consent and the ability to set their own terms, isn't a new demand. It's a restatement of a principle photographers won in court over 160 years ago, one Meta's product team either didn't consider or didn't consider a priority.

In 2004, Zuckerberg Shared His View on Ethics

In March 2004, a nineteen year old Mark Zuckerberg was messaging a friend during the early dispute over who owned the idea that became Facebook, the conflict that eventually produced the Winklevoss lawsuit. In that exchange, later documented by Harvard classmate Aaron Greenspan and republished in Ben Mezrich's 2019 Vanity Fair piece on the Winklevoss litigation, Zuckerberg wrote: "You can be unethical and still be legal; that's the way I live my life."

I want to be precise about what this quote is and isn't. It's a private message from a teenager in the middle of a personal dispute, not a corporate memo about AI policy, and it predates Muse by twenty two years. I'm not claiming he was describing this feature in advance. What I'm claiming is narrower and, I think, more useful: this is a documented statement of an operating principle from the person who has run this company for its entire existence, and the pattern of behavior since, including the Cambridge Analytica episode and now two AI products in two years that launched wide and retreated under pressure, is exactly what that principle predicts.

Law Is Downstream of Ethics,
Not the Other Way Around

Here's the relationship between the two events separated by 142 years. Law, especially around new technology, is almost always downstream of ethics. Someone identifies a harm, argues about it publicly, and only after enough consensus forms does a legislature or a court write it down. The 1862 ruling didn't invent the idea that a photographer owns their work. It formalized an ethical claim that photographers and their allies had already been making for years, over the objection of established artists who had every incentive to say otherwise.

Emerging technology creates a window before that formalization happens, and the length of that window is exactly where a company's actual values show up. You can spend the window asking what's right and building toward it. Or you can spend it asking what you can get away with until someone writes a rule against it, and treat the eventual apology as a cost of doing business, priced in from the start. Zuckerberg named the second approach as his personal operating principle two decades before he had an AI product capable of running it at this scale. Muse is what that principle looks like once the technology catches up to the philosophy.

2026 Legislation Is Catching Up to a 160 Year Old Principle

The window is closing, which is exactly what you'd expect after 160 years of the same argument repeating itself. In May 2026, the NO FAKES Act was reintroduced in the U.S. Senate and House with backing from the AFL-CIO, Google, IBM, the Motion Picture Association, OpenAI, and the RIAA, aimed at giving people a federal right against unauthorized AI replicas of their voice and likeness. That same month, the notice and removal provisions of the TAKE IT DOWN Act went into effect, requiring platforms to remove reported non-consensual intimate imagery, AI generated or not, within 48 hours. In June 2026, New York's FAIR News Act moved through both chambers of the state legislature, requiring disclosure when AI is used in published news content. SAG-AFTRA members that same month ratified new TV and theatrical contract terms specifically restricting synthetic replacement of performers' work.

None of this legislation is a reaction to some unprecedented new harm. It's the codification of an ethical argument that's been building since well before generative AI existed, the same argument Mayer et Pierson won in a French courtroom in 1862: the person in the image, or the person who made the image, gets to decide how it's used. The law is simply catching up to a position that was already the correct one to hold.

Photographers Already Own This Right.
Muse Tried to Take It by Default.

If you're a working photographer, this isn't an abstract debate about AI ethics. It's the exact right you've built your business on, being tested at platform scale by a company that has, by its own founder's account, treated legality and ethics as two separate and sometimes unrelated standards. The 1862 ruling and everything downstream of it, model releases, usage licenses, copyright registration, is the paperwork version of an ethical claim you already believe: your image, your terms. Muse tried to make that optional by default. It got walked back once enough people noticed and pushed back.

That's the part worth remembering the next time a platform launches something that quietly assumes your work is theirs until you say otherwise. The default they chose says more about their values than any apology that follows it.