Apple sued OpenAI over stolen secrets. The complaint says more about what Apple fears than about what OpenAI took.

On Friday, in the U.S. District Court for the Northern District of California, Apple filed a complaint that reads, on its surface, like the simplest kind of corporate grievance: someone took our things. The company alleges that OpenAI, "at every level, from members of its Technical Staff to its Chief Hardware Officer, and in coordination with business partners," has been stealing Apple's trade secrets and confidential information. It is a strong sentence. It is also, read the way a company's own words deserve to be read, a kind of confession — not of wrongdoing, but of fear.

Start with what the complaint actually alleges, because the specifics matter and most of the coverage stopped at the headline. The filing names two former Apple employees. Tang Tan spent twenty-four years at Apple, latterly as a vice president of product design responsible for the iPhone and the Apple Watch; he is now OpenAI's chief hardware officer. Apple says he used its confidential project code names while recruiting, asked candidates to bring Apple hardware components to their interviews, coached departing employees on how to get past Apple's security, and solicited details of unannounced products. The second, Chang Liu, an electrical engineer who spent eight years at Apple before leaving in 2026, allegedly failed to return a company laptop and used it to download confidential technical documents, which he then shared with other Apple employees who were interviewing at OpenAI.

Among the secrets Apple says were misused is a "proprietary metal finishing technique." The complaint also names io, the hardware startup OpenAI bought for a reported $6.5 billion in 2025. It does not name Jony Ive, io's co-founder. And it puts a number on the movement of people: more than four hundred former Apple employees, Apple says, now work at OpenAI.

What a lawsuit is for

A lawsuit is a document, and like every document a large company publishes, it is written to be read a particular way. Apple would like you to read this one as a matter of principle: theft is theft, and the company is defending property it built. That reading is not wrong. But it is not the whole document, and the more useful question — the one Apple's own choices answer — is why file it, and why now.

Companies that are winning a market do not usually sue the companies chasing them. They ship. Apple's most effective answer to a rival product has always been a better product of its own, delivered on Apple's timeline and at Apple's margin. The decision to reach instead for trade-secret law is a tell. It says the ordinary answer — out-build them — was not available fast enough.

Companies that are winning a market do not sue the companies chasing them. They ship.

Read the relief Apple asks for and the tell gets louder. The company is not, at this stage, chasing a dollar figure. It wants a court to bar OpenAI from using the information, to order the return of confidential materials, and to preserve evidence. Those are the remedies of a company that wants to slow something down, not one that wants to be paid.

• An injunction barring OpenAI from using the trade secrets Apple says it took.
• An order requiring the return of Apple's confidential materials.
• An order to preserve evidence — the documents, the devices, the communications.
• No specified damages figure, at least not yet.

Damages compensate for a loss already taken. An injunction is an attempt to change what happens next. Read the two together and the posture is clear: Apple is litigating the future, not the past. That is a strategic choice, and it is a different one from the choice the press release describes.

The crown jewels are boring

The detail everyone skipped is the one that gives the game away: a "proprietary metal finishing technique." It is not a moonshot algorithm or a secret chip. It is manufacturing know-how — the accumulated, unglamorous knowledge of how to make a metal object feel the way an Apple object feels, at a scale of hundreds of millions of units, without the yield collapsing. That is Apple's actual moat. Anyone can draw a beautiful device. Almost no one can mass-produce one.

So notice what the complaint implies about what OpenAI needs. A software company that has decided to make hardware does not lack for ideas or industrial design; it bought both when it bought io. What it lacks is the thing you cannot acquire in a deal — the process knowledge that lives in the heads of people who have shipped physical products at Apple's volume. Four hundred of those people now work down the road. The lawsuit, translated, is Apple's admission of where its real advantage sits, and its recognition that the advantage can walk out of the building.

The number that measures the wrong thing

Apple offers the four-hundred figure as evidence of a pattern — a coordinated raid. Read as a business fact rather than a legal one, it measures something less sinister and more uncomfortable for Apple: OpenAI is now where a particular kind of Apple engineer wants to work. Talent moves toward the project it believes in. Four hundred departures is not, by itself, evidence of theft. It is evidence of gravity.

The complaint treats the outflow as a crime scene. The more honest reading is a labor market voting with its feet, and voting against the idea that the most interesting hardware problem in consumer technology is still inside Apple. A company confident in its own gravitational pull does not usually count the people who left. It counts the ones lining up to join.

The device Apple won't name

The most revealing thing in the filing is the thing it is written to protect against: OpenAI's hardware. OpenAI has said little officially, but the shape has been widely reported — a consumer device built around an AI assistant rather than a grid of apps, developed out of the io acquisition and the team Apple says it was stripped for. Apple's complaint is, in effect, a description of a competitor's product that Apple takes seriously enough to fight in court before it has even shipped. You do not sue over a toy.

This is the part a keynote will never say. On stage, Apple's posture toward artificial intelligence has been that it is being careful — that it will arrive when the experience is right. That framing has worn less well over the past year, as the company's own assistant work reportedly slipped and leaned on outside models. The lawsuit is the counter-document. It concedes, in the dry language of tort, that a rival's device is far enough along, and close enough to Apple's own playbook, to be worth the cost and the spectacle of suing. Confidence does not litigate. Anxiety does.

The strongest version of Apple's case

None of this means the complaint is baseless, and it is worth stating the strongest version of Apple's case, because part of it is real. An engineer who keeps a company laptop and downloads documents on his way out is not a labor-market abstraction. That is the specific, individual act trade-secret law was written for, and if Apple can prove that particular conduct, it may well win on it. Misappropriation is a real wrong, and "everyone does it" has never been a defense.

But winning on the laptop is a narrow victory, and Apple's complaint is not written narrowly. It is written to describe a scheme — a coordinated effort running from junior staff up to a chief officer — because the scheme is the point. And a scheme is much harder to prove than a laptop. The gap between those two documents, the provable misconduct and the alleged campaign, is exactly where the strategy lives. Apple does not need to win the campaign in the end. It needs the campaign alleged now, on the docket, attached to discovery and depositions and the slow friction of litigation, wrapped around the precise team that is trying to ship a device on a schedule.

The timing, and the sentence OpenAI chose

The suit lands in a loud week. It arrived as OpenAI was pressing its enterprise ambitions and as its leadership traded public shots with rivals — the kind of noise that makes a legal filing read as one more front in a personality war. Set the noise aside. The filing is not a post on social media; it is a document Apple's lawyers will have to stand behind for years. Its timing says Apple decided the cost of waiting — of letting the device get closer to launch without a cloud over it — was higher than the cost of filing early and loud.

OpenAI's answer was a single clean sentence: "We have no interest in other companies' trade secrets. We remain focused on building innovative technology that empowers people everywhere." It is a good sentence, and it is doing its own careful work. It denies wanting secrets in the abstract without addressing the specific laptop, the specific downloads, or the specific hires. Both companies are choosing their words carefully this week. Only one of them filed a lawsuit's worth of them.

What Apple is actually buying

So read the complaint as what it is: a purchase. Apple is spending legal fees and a measure of dignity — the company that sells itself as the inventor of the future is, this week, the one asking a court to slow someone else's — to buy time and friction against a hardware program it cannot yet outpace on the merits. That may be a shrewd purchase. Litigation is a tool, and using it to protect genuine process secrets is legitimate. But it is a tool that works on the court's clock, not Apple's, and reaching for it is itself information.

The real strategy is not written anywhere in the filing, because companies never write it down. It is this: Apple has concluded that OpenAI's device is a genuine threat, that it cannot beat it to market on the current schedule, and that the instruments it has left are the ones that run on someone else's schedule instead of its own. The lawsuit is not a sign of Apple's strength in AI hardware. It is the most detailed thing the company has ever published about the limits of it. When a company reaches for the law instead of the product, the law is the part it is confident about.