Apple employees could leave for OpenAI; Apple’s trade secrets, the company argues, could not leave with them. In December 2023, that distinction was invisible inside an ordinary personnel story. Roughly two and a half years later, it is the subject of a lawsuit.

People were the first unit of competition

In December 2023, sources said Tang Tan planned to leave Apple in February 2024. Tan was Apple’s vice president for iPhone and Watch product design, which meant the departure involved more than a recognizable executive: it moved accumulated experience in turning design decisions into consumer hardware.

At that stage, the departure still fit the familiar structure of technology competition. A company loses an executive; another effort gains one. Knowledge moves because people move, and the boundary between personal experience and company property remains mostly invisible while nobody asks a court to draw it.

Each subsequent phase made the hardware effort more concrete:

The sequence matters because no single departure produced the reversal. First came an executive move, then a named venture, then an OpenAI hardware organization capable of recruiting additional Apple engineers. What looked like a talent pipeline began to look, from Apple’s side, like a channel through which proprietary knowledge might travel.

Compensation made recruitment part of the infrastructure

The economics surrounding that effort were exceptional. According to Equilar, OpenAI’s companywide stock-based compensation in 2025 averaged:

per employee, about 34× the pre-IPO peer average

The figure is not specific to hardware and says nothing about misconduct. It does show the conditions surrounding the contest: OpenAI built a compensation system capable of assigning extraordinary value to scarce people. Its average was roughly seven times Google’s level before its IPO.

That system creates a reinforcing loop. Rich equity helps attract experienced employees; experienced employees reduce the distance between an ambition and an operating organization. The resulting organization becomes more credible, making its equity and mission more attractive to the next recruit. The loop works well while talent is treated as a portable input.

Hardware eventually exposes the limit of that assumption. A consumer device is not an org chart assembled from impressive résumés. It is accumulated judgment about design, engineering and production, some carried as experience and some claimed by an employer as confidential. Recruitment moves the first category by design. Apple’s lawsuit alleges that OpenAI’s effort reached into the second.

The boundary became valuable because hiring worked so well.

The scarce asset shifted from talent to organizational memory

When a new field begins, hiring often stands in for capability. Senior recruits and former employers suggest what a new organization can build. But that proxy breaks once the organization has enough talent to pursue a consumer device. The relevant question is no longer who joined. It is what those people know and which parts they may use. The hard part is distinguishing retained expertise from protected information after both enter the same building.

Apple says iPhone engineer Chang Liu left for a job in OpenAI’s nascent hardware division with more than his years of experience. The lawsuit follows months of tension around OpenAI’s hardware effort and Tan’s relationship with his former employer. Apple separately alleges that Tan directed Apple employees interviewing at OpenAI to share company secrets.

Those claims remain unproven. Nor does one dispute establish that the entire AI-hardware sector has become more litigious. The narrower structural shift is already visible: in the Apple-OpenAI branch of the race, personnel movement is no longer treated merely as competition for labor. It is being contested as a possible transfer mechanism for proprietary capability.

The hiring strategy has become the legal exposure

This is the reversal. The strategy was to accelerate hardware development by recruiting people who already understood consumer devices. That experience made the recruits valuable. It also made their movement legally sensitive, because the receiving organization must show where expertise ends and protected information begins.

The actors matter to the case, but the pressure is not personal. Replace Tan with another hardware leader, Apple with another integrated device maker, or OpenAI with another model company attempting to become a product company, and the design problem remains. A hiring system built to import capability eventually imports disputes over the ownership of that capability.

For Apple, defending trade secrets means keeping proprietary knowledge inside the institution even when employees leave. For OpenAI, recruiting veteran hardware personnel means making knowledge usable in a new institution. Both systems can behave as designed until the same engineer sits at their boundary. Then retention and recruitment become opposing claims on what one person carries.

The first headline was about a vice president scheduled to leave in February. The present case is about what followed other employees into OpenAI’s hardware division. The résumé crossed the door; the lawsuit is over whether the blueprint crossed with it.