OpenAI’s Sudden Pledge on Apple Secrets Exposes Raw Tensions in Hardware Race

 

In a San Jose federal courtroom this week, OpenAI promised not to touch any Apple trade secrets its former employees might have carried over. The concession came moments after U.S. District Judge Edward Davila signaled he was highly likely to issue the injunction Apple had demanded.

Hours earlier, OpenAI had turned the accusations around. Its own filing accused an unnamed ex-employee, who had joined from Apple, spent six months at the AI company, then returned to Cupertino, of sending more than 100 highly confidential files about an unreleased OpenAI device to personal accounts. Some transfers happened just after midnight in Singapore on the employee’s final day in June 2026.

The documents included detailed images of the hardware, engineering specifications, manufacturing data, supplier details and standards for cosmetic finishes. OpenAI told the court it believes the former staffer later presented those product plans to Apple’s senior leadership, including the vice president of operations in Asia. “OpenAI is very concerned that its trade secret and confidential information is now circulating inside Apple, where Apple engineers can shape Apple’s upcoming products based on OpenAI’s confidential and innovative work,” the filing stated, according to Reuters.

Apple pushed back immediately. It said it maintains strict policies to prevent new hires from bringing in material from previous employers. The company added it had found no evidence supporting OpenAI’s claims and would cooperate to remove any such data if located. Those details emerged in coverage from CNN Business.

This back-and-forth marks the latest chapter in a lawsuit Apple filed in July 2026 against OpenAI, its hardware-focused unit io Products, and two former Apple engineers: Chang Liu, a senior systems electrical engineer, and Tang Yew Tan, now OpenAI’s chief hardware officer. Apple alleges a coordinated effort to extract proprietary information on unreleased products, manufacturing processes, components, testing methods and supplier relationships. The goal, Apple claims, was to accelerate OpenAI’s push into consumer hardware. An AppleInsider report captured the unusual tone of OpenAI’s response, noting the company appeared to admit possession while swearing it would not use the material.

Yet the case runs deeper than one employee’s actions. On October 13, Judge Davila granted Apple much of its request for expedited discovery. OpenAI must turn over forensic images already collected from relevant devices, including those examined by its own experts. It must produce interview guides, checklists and technical presentations used when speaking with Apple recruits for its consumer hardware group. And it must hand over LINE messages exchanged between Liu and Yu-Ting “Alyssa” Peng since January 1, 2026. Those chats reportedly included discussions about restoring and then actively using Apple-issued devices, a point the judge cited as evidence of risk that critical data could vanish. Bloomberg Law reported the scope of that order.

Depositions were approved on an accelerated schedule for Liu, Tan and Peng, plus corporate representatives from OpenAI and io Products. The judge stopped short of giving Apple everything. Requests for broad access to materials on metal finishes, power management and battery research were deferred to the normal timeline. Still, the partial win gave Apple early momentum.

OpenAI’s hardware ambitions sit at the center. The company acquired io Products, the startup founded by former Apple design chief Jony Ive, in a $6.4 billion deal last year. That move signaled serious intent to build physical products powered by its AI models. The device it is working on is reported to be a small screen-free smart speaker. But building hardware demands expertise that Apple has spent decades refining. Apple says OpenAI has hired more than 400 of its former employees. The question is whether that talent brought forbidden knowledge along.

Both sides now face forensic scrutiny. OpenAI agreed to hand over forensic images of the drives of 14 employees who previously worked at Apple. The material will be preserved while the case continues. Judge Davila can enforce the no-use pledge. The enforceable commitment avoids the immediate injunction yet binds OpenAI’s development teams. One misstep could bring contempt findings.

The talent war between these organizations has grown fierce. OpenAI has argued Apple’s suit amounts to sour grapes over losing key people in a competitive market. It has denied the allegations. Apple, meanwhile, portrays a pattern. Its complaint described former employees downloading confidential files and urging job candidates to bring Apple parts to interviews at OpenAI.

Recent coverage shows the claims have only sharpened. The Verge noted OpenAI’s concession followed the judge’s strong indication he would side with Apple on the preliminary injunction. That report, published October 15, highlighted how the voluntary agreement lets OpenAI avoid a formal court order while still facing the underlying allegations.

Tech executives on both sides understand the stakes. Consumer AI hardware represents the next major battleground after software and cloud services. Voice interfaces, on-device intelligence and personal robotics all require tight integration of silicon, sensors, batteries and manufacturing scale. Companies that master those combinations first could lock in decades of advantage. Yet protecting the underlying knowledge grows harder as engineers move between firms carrying years of hard-won experience in their heads.

Trade secret law draws a line between general skills and specific proprietary information. Courts often struggle to separate the two. Judge Davila’s comments suggest he sees enough smoke to justify early evidence gathering. The forensic images, messages and interview notes should clarify whether specific Apple files ended up shaping OpenAI’s designs.

And the counter-accusation adds symmetry. If OpenAI’s claims hold, Apple now sits in the uncomfortable position of defending against similar charges. Its response has been measured: it has policies in place and says it has found no evidence. That stance buys time but won’t satisfy OpenAI if the files surface inside its walls.

Observers following the case expect months of discovery battles ahead. Depositions of the named individuals could prove revealing. So could the LINE chats. Those private messages, referenced by both parties, may contain the unguarded language that tips the scale. For now the litigation has produced an uneasy truce on one set of secrets while opening questions about another.