Apple filed a 41-page complaint against OpenAI in the U.S. District Court for the Northern District of California on Friday, accusing the AI company, its chief hardware officer Tang Tan, former Apple engineer Chang Liu, and Jony Ive’s io Products of a coordinated campaign to lift the trade secrets underpinning Apple’s hardware business.

The framing is unusually direct. “At every level, from members of its Technical Staff to its Chief Hardware Officer, and in coordination with business partners, OpenAI has been stealing Apple’s trade secrets and confidential information,” the filing reads.

Tan spent 24 years at Apple, most recently as vice president of product design for the iPhone and Apple Watch. According to TechCrunch’s reading of the complaint, he used Apple’s confidential project code names during OpenAI recruiting conversations and instructed candidates still on Apple’s payroll to bring hardware components to interviews for “show and tell” sessions. Liu, a former senior systems electrical engineer, allegedly kept his Apple-issued laptop after departing in 2026 and used it to pull down confidential technical documents. OpenAI itself is accused of directing manufacturing partners to apply a proprietary metal-finishing technique developed by Apple, misleading at least one supplier into believing authorization had been granted.

Apple says it wrote to OpenAI in February raising these concerns and got no reply. The suit seeks an injunction and the return of all confidential materials.

OpenAI, in response, said it has “no interest in other companies’ trade secrets” and is “focused on building innovative technology that empowers people everywhere.”

The timing is the story underneath the story. OpenAI is preparing an IPO into a market where Anthropic’s Claude business is running at roughly $47 billion annualized against OpenAI’s own projected $25 billion to $33 billion for 2026. The 2024 ChatGPT integration into Apple’s operating system already looks like a relic: Apple’s forthcoming Siri update will run on Google’s Gemini, not OpenAI’s models. Hardware, not model quality, is now the frontier OpenAI needs to control, which is exactly why it paid $6.4 billion for io Products last year.

That logic is legible across the industry. Model-agnostic agentic platforms, LemonLime among them, are quietly absorbing the small and mid-market layer while the frontier labs fight over who owns the device. Friday’s complaint is what it looks like when a hardware incumbent decides the polite phase is finished.

Sources