In December 2023, Tang Tan’s reported departure from Apple was a personnel story. Roughly two and a half years later, it sits inside Apple’s trade-secret lawsuit against OpenAI.

A departure is small until the departures interlock

Corporate departures are usually counted one employee at a time: an executive leaves, the incumbent reorganizes, and the move becomes a line in a personnel announcement rather than a change in industrial structure.

Tan’s exit initially fit that pattern. The story had one name, one role, and one destination. Apple could replace a role.

Then the names began to form a system.

None of that sequence, alone or together, proves misconduct. It establishes something more basic: the unit of competition changed. Tan was no longer merely a former Apple executive pursuing a new project. He had become OpenAI’s hardware chief, while a cluster of people associated with Apple’s product and design system had gathered around an AI-native consumer-device effort.

The talent market moved more than labor

Hardware expertise is difficult to contain because its most valuable parts are not always reducible to a single drawing or file. They include the sequencing of decisions: which constraint appears first, which compromise can survive manufacturing, which product choice will force changes elsewhere. A company can own documents. The people who learned how those documents became devices still retain experience.

Mobility across that boundary is normally productive. Moving people moves capability, and moving capability lets new organizations form. But the mechanism contains a reinforcing loop: credible leaders attract specialists, the concentration of specialists makes the project more credible, and that credibility attracts more people. Each hire looks ordinary until the accumulation crosses a threshold.

OpenAI’s average stock-based compensation per employee in 2025, according to Equilar—roughly seven times Google’s pre-IPO level and about 34 times the average of other pre-IPO peers.

That figure does not reveal what any hardware recruit received, and it proves nothing about how anyone was hired. It does describe the economic field around the migration. OpenAI was not recruiting with the incentives of an ordinary young company; it had the capacity to make an established institution’s accumulated experience unusually mobile.

That combination produced a structural reversal. Apple’s integrated product system developed and concentrated people capable of building integrated products. Once enough of those people assembled elsewhere, the same system became a source of competitive exposure. The risk was not that every departure was improper. It was that separate departures could recombine into an external production system.

The lawsuit changes the unit from people to information

Apple’s complaint marks the threshold because it no longer treats the movement as a series of hires. According to Apple’s allegations, iPhone engineer Chang Liu left for OpenAI’s nascent hardware division carrying more than his lawful experience. Apple also alleges that Tan directed Apple employees interviewing at OpenAI to share company secrets.

Those are allegations, not adjudicated findings. The evidence of an Apple-to-AI-hardware talent pipeline does not establish that every hire, or even most hires, involved improper conduct. Ordinary recruitment and trade-secret theft are not interchangeable simply because they occur along the same route.

That distinction locates the structural shift. Experience and institutional knowledge travel together until trade-secret law must draw a boundary between them. What looked like labor mobility has become, from Apple’s perspective, a channel through which a rival may acquire protected know-how.

Reports have also emphasized strain between Tan and his former boss, Ternus. That makes the story easier to personalize and harder to understand. The actors could be different and the pressure would remain: OpenAI needs hardware capability it did not spend decades developing, while Apple needs to preserve the product knowledge embedded in employees who can legally resign.

Ternus’s position makes the institutional response visible. His appointment came amid historic turnover and a stated challenge to retain top talent. Reports that he plans to reshape industrial design and reassert its executive influence describe one response: strengthen the center. The lawsuit describes another: reinforce the boundary.

Software can advance while hardware hardens

OpenAI’s hardware push is not replacing its consumer-software strategy. On the same day the legal conflict came into view, the company broadly released GPT-5.6 and launched ChatGPT Work for macOS and Windows. The software cadence continued alongside the dispute.

The simultaneity separates two layers of the same system. Software can spread through computers designed and controlled by existing platform companies; an AI-native consumer device requires a separate physical product and the organizational capability to make it. OpenAI can keep shipping software while assembling that capability, and Apple can remain a platform for OpenAI’s applications while contesting how its hardware knowledge reached OpenAI’s organization.

In December 2023, Tang Tan was one name on an exit notice. Roughly two and a half years later, Apple is asking a court to draw a line between engineers’ portable expertise and its protected knowledge. The notice was paperwork; stacked with the others, it became the blueprint Apple says walked out the door.