Apple’s OpenAI lawsuit turns the AI hardware race into a trade-secret fight

A partnership meets a lawsuit

Apple and OpenAI already have a complicated relationship. Apple uses OpenAI technology in parts of its device software. At the same time, both companies are trying to shape what comes after today’s phone-and-app model of computing.

That relationship now includes a federal lawsuit.

On 10 July, Apple filed a civil complaint in the Northern District of California against OpenAI, several related entities including io Products, and two former Apple employees: Chang Liu and Tang Yew Tan. The court docket lists trade-secret misappropriation under the federal Defend Trade Secrets Act, along with breach-of-contract claims. Apple has asked for a jury trial.

The filing is a serious escalation. It is not a verdict. At this stage, the public record contains Apple’s account of what happened, not findings by a judge or jury.

What Apple alleges

Apple’s complaint centres on the movement of people from its hardware organisation into OpenAI’s expanding device effort.

According to the complaint as reported by the Associated Press, Apple alleges that Liu kept an Apple-issued laptop after leaving the company and accessed or downloaded confidential material. Apple also accuses Tan, who now leads hardware work at OpenAI, of encouraging recruits to bring detailed knowledge and even physical parts from Apple projects.

Apple’s broader allegation is that OpenAI did not merely hire experienced people. It says the company benefited from information that remained Apple’s property, including aspects of product design, manufacturing and supplier relationships.

Those details come from Apple’s complaint. They should be read as claims that Apple will have to support with evidence. The defendants will have an opportunity to dispute both the facts and Apple’s legal interpretation of them.

What OpenAI says

OpenAI has not, at the time of publication, published a detailed response to each allegation. A spokesperson told the Associated Press that the company has no interest in other companies’ trade secrets and was still reviewing the filing.

That response rejects the premise of deliberate misuse without yet addressing the complaint point by point. The next useful evidence is likely to come through the case itself: formal answers, motions, document production and, if the dispute continues, testimony.

Apple has likewise made a public statement that it intends to defend the work and innovations of its teams. Neither statement resolves the factual dispute.

Why hardware makes the boundary harder to police

People carry expertise when they change jobs. An engineer does not forget how to solve a thermal problem or organise a difficult manufacturing programme. Companies are allowed to hire that experience, and workers are generally allowed to use the skills they have developed.

Trade-secret law draws a different line around information that is valuable because it is not public and that a company has taken reasonable steps to protect. Internal specifications, unreleased prototypes, supplier terms and confidential test results can fall on that side of the line.

The distinction sounds neat in a legal definition. It is much harder inside a fast-moving product team. A useful idea may sit in someone’s memory, a private document, an old message or a physical component. A new employer therefore needs more than a clause in an offer letter. It needs practical controls: clear instructions, clean-room processes for sensitive work, limits on imported files and a culture in which speed does not excuse ambiguity.

The AI hardware race raises the stakes

AI companies increasingly want a direct relationship with the person using their models. Hardware offers that possibility. A device can control microphones, cameras, sensors, battery life and the moments when an assistant is available. It can also create dependence on difficult manufacturing and supply chains that established consumer-electronics companies have spent decades building.

That makes experienced hardware staff unusually valuable. It also makes disputes over what those people know more likely.

The Apple case will not settle every boundary. Its immediate question is narrower: whether specific protected information was taken, shared or used, and whether the companies and individuals named in the complaint are legally responsible. But the litigation may reveal how one of the world’s best-funded AI hardware efforts recruited its team and tried to keep old and new work separate.

What to watch next

The most important next step is the defendants’ formal response. It should show which allegations are admitted, denied or challenged as legally insufficient. The court may also have to decide how much of the evidence can be public, since litigation about trade secrets often involves material that the parties say must remain confidential.

Until then, dramatic language should be treated carefully. Apple has documented a claim; it has not proved theft. OpenAI has issued a denial in principle; it has not yet supplied a detailed public rebuttal.

The useful lesson for the wider industry is already visible. The race to build AI devices is also a race for scarce human expertise. The companies moving fastest will still need to show that they know where expertise ends and somebody else’s confidential property begins.