Apple sues OpenAI over hardware trade secret theft

The most extraordinary allegations in Apple's complaint are not about a disgruntled employee taking files home - they describe an institutionalized acquisition program. Tang Yew Tan, OpenAI's Chief Hardware Officer and a 24-year Apple veteran, is accused of directing job candidates to bring physical hardware parts and CAD artifacts to OpenAI interviews for 'show and tell' sessions ^[1]. He is also alleged to have coached departing Apple employees on how to evade Apple's own exit security procedures - using Apple's internal 'Need to Know' security documents as the coaching material ^[1].

Chang Liu's alleged conduct is equally brazen. After leaving Apple in January 2026, Liu reportedly exploited a network authentication bug that left his credentials active post-departure, accessed Apple's internal cloud file storage, and downloaded over 1,000 pages of confidential technical files. His alleged message to a colleague at the time: 'LOL, I found out I can access the [network storage], so funny' ^[2].

Beyond individuals, the complaint alleges corporate-level appropriation: OpenAI and io Products allegedly obtained Apple's proprietary metal-finishing technique by misleading an Apple supplier partner into believing OpenAI had Apple's permission to use it, and used Apple's own internal supply chain terminology with vendors ^[3]. Apple's complaint states: 'This is the tip of the iceberg. Apple lacks visibility into what's happening behind closed doors at OpenAI, where such misconduct is normalized and exemplified by leadership.' The firm notes that 400+ former Apple employees now work at OpenAI - a pipeline that allegedly created systemic exposure across product, manufacturing, and supply chain domains ^[4].

The timing of Apple's lawsuit is surgically precise. OpenAI's first consumer hardware product - codenamed 'Gumdrop,' a portable screenless smart speaker co-designed with Jony Ive - was slated for announcement in 2026 and release in 2027 ^[5]. Apple is seeking an injunction that would require OpenAI to identify and isolate all disputed materials and certify compliance. Bloomberg Intelligence assesses Apple is likely to secure targeted preliminary relief - and even a partial injunction forces a product redesign audit that materially slows the timeline ^[6].

The IPO consequences may be more immediate. OpenAI, valued at $852 billion as of April 2026, now faces mandatory material-risk disclosure in any IPO prospectus covering the lawsuit and potential injunctive outcomes ^[7]. Prediction markets registered the shift within days: the probability of a 2026 IPO fell from approximately 22% to 18.5% after the filing. OpenAI also faces concurrent investigations from 42 state attorneys general, compounding the legal overhead heading into public markets ^[7].

Personal naming of defendants Tang Tan and Chang Liu sends a secondary signal with its own strategic effect: any Apple employee now considering joining OpenAI faces explicit personal legal exposure as a deterrent. Apple's complaint cover language frames the case in maximalist terms - 'OpenAI's nascent hardware business now rests on the shakiest of foundations, rotten to its core by its illegal reliance on misappropriated trade secrets' - language calibrated less for a judge than for OpenAI's future recruits and investors ^[3].

The Waymo v. Uber trade secret case offers the closest legal parallel, and the comparison is structurally apt. In the Waymo case, Uber's self-driving division head Anthony Levandowski departed Google, allegedly taking thousands of proprietary files, and founded a startup that Uber then acquired. The outcome: Uber was forced to surrender equity and submit to years of audit requirements; its self-driving division ultimately shut down.

The Apple-OpenAI case has features that could make it more severe. First, the alleged conduct was not isolated to one employee: Apple's complaint names actors at every organizational level, from technical staff to the Chief Hardware Officer, and alleges that Apple's own security documents were weaponized to systematize the theft ^[3]. Second, the alleged misappropriation extends beyond digital files to physical hardware, supplier relationships, and proprietary manufacturing processes - making a clean 'return the files' remedy structurally inadequate. Third, the sheer scale of the alleged pipeline - 400+ former Apple employees - means discovery could surface evidence well beyond the named defendants. Apple's complaint explicitly anticipates this: 'Discovery will expose that the misappropriation has been occurring on a scale many times greater than the several instances described below' ^[3].

California's hostility to non-compete agreements will limit some of Apple's legal tools, but the physical hardware and supplier deception allegations go well beyond standard NDA claims into federal trade secret territory - where remedies are broader and the threshold for injunctive relief is lower.

Bloomberg Opinion offered the most non-obvious take on the filing: Apple's lawsuit may turn out to be a favor to OpenAI ^[8]. The argument runs as follows. OpenAI is burning cash at a historic pace ahead of its IPO. The Gumdrop device would place it in direct competition with Apple - the world's most profitable consumer hardware company, with decades of manufacturing relationships, retail infrastructure, and supply chain leverage that no amount of trade secret acquisition can replicate overnight. A court-ordered injunction, or the credible threat of one, gives OpenAI's board political cover to pause or restructure the hardware effort without appearing to retreat.

The $6.5 billion acquisition of io Products was always a bet with asymmetric risk: the upside was owning the consumer interface layer in an AI-native device era; the downside was a multi-year hardware development cycle against a competitor that invented the category. The lawsuit crystallizes that downside into a legal document Apple's own lawyers will pursue aggressively. Sam Altman's public posture - 'not afraid of Apple' - is the only statement he can make heading into IPO roadshows, but it does not reflect the structural reality that a preliminary injunction forcing hardware redesign would push the Gumdrop launch well past 2027 ^[5].

For the broader AI industry, this case sets a legal standard that will shape how AI companies recruit from hardware incumbents for years. Paolo Pescatore's observation that the AI race is now being fought in factories and living rooms, not just data centers, captures why the stakes extend beyond these two companies ^[9]. The legal template Apple is building here will be cited in the next dozen cases where an AI company poaches engineering talent from a hardware incumbent.