On February 23, 2026, Gabriel Gross, Apple's outside counsel at Weil, Gotshal & Manges, sent an email to Che Chang, OpenAI's General Counsel. The subject line read: "Former Apple Employees at OpenAI Retaining Non-public, Confidential, and Proprietary Information." The message stated a wish to discuss the possibility that former Apple employees who had moved to OpenAI were retaining confidential information.
But at 6:07 PM that same day, Gross sent another message to the same email thread by mistake. It had been intended for a different former Apple employee, "Wang," but was accidentally forwarded to Chang instead. Gross later acknowledged the error himself, explaining to Chang that "Mr. Wang immediately offered to cooperate." According to OpenAI's blog post published August 3, titled "Apple is getting this wrong," Apple's in-house counsel confirmed this mix-up two days later, but only conveyed that "Weil is representing us on this matter." No specifics of any claim were ever presented.
For the following five months, Apple went silent — no further contact with OpenAI. Then, on July 10, Apple filed a complaint in the U.S. District Court for the Northern District of California (Case No. 5:26-cv-07078).
Apple's Narrative of "Organized Theft"
Apple's complaint is a 41-page document naming two former Apple employees, OpenAI, and OpenAI's hardware division, io Products, as defendants. It opens: "This case is about former Apple employees stealing Apple's trade secrets and using them for OpenAI's benefit."
One of the defendants, Chang Liu, spent eight years as a Senior Systems Electrical Engineer at Apple before departing in January 2026 to join OpenAI's technical team. According to the complaint, Liu failed to return his Apple-issued laptop upon leaving, and later exploited a discovered authentication bug to access Apple's shared network folders, downloading dozens of confidential files containing "extensive details regarding unreleased products, engineering presentations, and technical specifications." Apple cites messages Liu allegedly left for colleagues — "LOL" and "so funny" — as evidence that he was "enjoying" the unauthorized access.
The other defendant, Tang Yew Tan, spent more than 24 years at Apple and served as Vice President of Product Design for iPhone and Apple Watch. He now serves as OpenAI's Chief Hardware Officer (CHO). The complaint alleges that Tan used internal Apple project codenames to extract information about unreleased products, had job candidates bring "actual parts" from Apple for "show and tell" sessions during interviews, and instructed new hires "not to tell Apple" they were moving to OpenAI.
The complaint offers one striking figure: OpenAI currently employs more than 400 former Apple employees.
OpenAI's Counterattack Was a Blog Post, Not a Court Filing
Late on August 3, OpenAI published a post titled "Apple is getting this wrong" on its official blog. It is not a legal answer. It is not a document filed with the court. It is a rebuttal aimed at the general public, complete with verbatim excerpts of emails and iMessages.
OpenAI's argument breaks down into three main points.
First, that Apple's claim in the complaint — that it reached out in February and OpenAI ignored it — is not accurate. According to the emails OpenAI published, Apple's counsel confused two Asian surnames (Chang and Wang) in a misaddressed email, falsely claimed to have "spoken by phone" with OpenAI's General Counsel, and then retracted that claim. No specifics of any legal claim were ever presented; Apple's side conveyed only that it was "resolving any issues."
Second, regarding the claim that Liu accessed Apple's confidential information after leaving the company. The iMessage screenshots OpenAI published show that even after Liu's last day of employment (January 22, 2026), current Apple employees were messaging Liu asking him to point them to the location of certain files. OpenAI argues that what Liu accessed was in response to requests from Apple employees themselves — not the "theft" Apple alleges — and that this instead reflects Apple's own failure to properly deactivate departed employees' access privileges.
Third, regarding Tan. OpenAI states that "Tang has always made clear to the team that even if they want another company's confidential information, they must not use it," and calls a man who devoted more than 24 years to Apple "one of the most innovative leaders" it knows.
At the opening of its blog post, OpenAI praised Apple as "one of the greatest companies of all time," while dismissing the lawsuit itself as "careless, aggressive, and oddly personal."
The Structure of a Battle Fought in the Court of Public Opinion
OpenAI's move opens a front separate from the legal proceedings themselves. A party to litigation publishing verbatim, out-of-court private communications with opposing counsel carries legal risk. The published correspondence could give the other side material for rebuttal, and could potentially be used against OpenAI itself as unfavorable evidence in court.
Still, OpenAI chose this path — and the reasoning lies in an asymmetry of timelines. The preliminary injunction Apple has requested is a measure that restricts defendants' conduct without waiting for a trial on the merits. If granted, Liu, Tan, and OpenAI would be barred from accessing, obtaining, using, or disclosing what Apple claims to be "confidential information" for the months or years the litigation continues. According to Reuters, Apple has simultaneously requested expedited discovery, seeking depositions from Liu, Tan, OpenAI employee Yu-Ting Peng, and another unnamed former Apple employee.
For OpenAI, if the injunction is granted, it would jeopardize the launch of its first hardware device, planned for later in 2026. According to Axios reporting from January, OpenAI's Chris Lehane (Chief Global Affairs Officer) had said the company was "on track to unveil our first device in the second half of 2026." Fortune reported in July that the device's design had already been finalized, and the io team had swelled to more than 400 people. The device is being developed by a team led by Jony Ive, stemming from OpenAI's May 2025 acquisition of io Products for $6.5 billion in an all-stock deal, based on OpenAI's $300 billion valuation at the time.
In other words, if the injunction is granted within the next few weeks, OpenAI's entire hardware strategy could be legally frozen. Waiting for a final verdict in the trial would be too late. That is precisely why OpenAI needed to undermine the credibility of Apple's claims in the court of public opinion before the court itself renders a decision.
The Unresolved Question Left by Waymo v. Uber
This type of dispute is not new. In 2017, Google's self-driving division Waymo sued former engineer Anthony Levandowski, alleging he downloaded more than 14,000 confidential files before moving to Uber. The case ended in a 2018 settlement in which Uber paid Waymo stock valued at $245 million — but no verdict was ever rendered.
The biggest unresolved question left by the Waymo case is whether "negative trade secrets" — knowledge of what does not work — are also protected as trade secrets. California prohibits noncompete agreements and does not recognize the "inevitable disclosure" doctrine. Employees are, in principle, free to use the general skills and knowledge they gained at a previous employer in their next job — but taking specific confidential documents is a different matter entirely. Where exactly is that boundary? Because the Waymo case ended in settlement, it never established binding case law.
The Apple v. OpenAI lawsuit raises this question anew. Is the manufacturing know-how, supplier relationships, and design philosophy Tan accumulated over 24 years "general skill," or is it a "trade secret"? Was Liu's act of telling Apple employees where to find files at their request "cooperation," or "unauthorized access"?
| Issue | Apple's Claim | OpenAI's Rebuttal |
|---|---|---|
| February contact | Contacted OpenAI, but was ignored | Counsel sent the message to the wrong person; no specific claims were ever presented |
| Liu's access | Exploited an authentication bug to steal confidential files | Only responded to requests from Apple employees. The access control failure is Apple's own problem |
| Tan's conduct | Used internal codenames to extract info on unreleased products; had candidates bring real parts | Tan has always instructed his team never to use another company's confidential information |
| Organized involvement | More than 400 former Apple employees now at OpenAI, systematically bringing in information | Movement of talent is free, and OpenAI holds no Apple trade secrets |
What the Published Communications Say — and What They Don't
When reading OpenAI's blog post, one thing must be kept in mind: the communications published were selected by OpenAI. Rather than complete email threads, only the portions that support OpenAI's claims were extracted. As The Verge points out, these are "cherry-picked communications," fundamentally different in nature from documents both sides submit through the formal discovery process.
For instance, the messages from Liu that Apple's complaint cites — "I still have another computer," "LOL," "so funny" — do not appear anywhere in OpenAI's blog post. What OpenAI published instead were iMessages of Apple employees asking Liu for help, scenes of Liu responding to Apple's own requests. From the same person's communications over the same period, which facets get selected can flip the story 180 degrees.
What rebuttal Apple will offer regarding the published communications remains to be seen. Apple had not issued any comment to The Verge at the time of that outlet's reporting.
Two Clocks: The Courtroom and Public Opinion
A ruling on a preliminary injunction typically comes within a few weeks. If Apple's request for expedited discovery is granted, the timeline for OpenAI's internal documents to reach Apple's hands would also be significantly accelerated.
Saurabh Vishnubhakat, an intellectual property law professor at Cardozo School of Law, told Fortune that this lawsuit is more likely to "delay, fragment, and force a rebuild" of OpenAI's hardware project than to "kill" it outright — potentially requiring the isolation of files and employees, redoing engineering work, and changing manufacturing partners.
The tactic OpenAI chose through its blog post was to dismantle Apple's narrative outside the courtroom, before the legal process could run its course. Whether it succeeds will depend on how the published communications are treated in court, and which way the ruling on the injunction ultimately tips. The clock of the courtroom and the clock of public opinion do not run at the same speed.