Apple’s OpenAI Injunction Bid Collides With Its Own Offboarding Mess

Apple asked a federal judge late Monday for a preliminary injunction barring OpenAI and two former Apple employees from accessing, using or sharing what it calls stolen hardware trade secrets, warning of irreparable harm if the ChatGPT maker’s device work continues unchecked.

Hours later OpenAI published OpenAI’s full blog response with emails and iMessages that cast the dispute as Cupertino’s own residual access failures and a lawyer who mixed up two Asian surnames. The filing lands in the Northern District of California as both sides race for the next consumer AI interface.

The clash joins a courtroom fight over people and files to a product fight over who reaches users first. Apple wants the court to lock down alleged secrets before any device work hardens into shipping hardware. OpenAI wants the public record to show process failures on Apple’s side and no appetite for Cupertino’s designs.

What Apple Asked the Court to Freeze

The motion seeks to stop Chang Liu, Tang Yew Tan and OpenAI from possessing, acquiring, using or disclosing alleged confidential information. Apple also demanded expedited discovery, including depositions of Liu, Tan, OpenAI staffer Yu-Ting Peng and an unnamed fourth former Apple employee now at the startup, plus corporate representatives of OpenAI and io Products.

• Core ban: no further access, use or sharing of claimed secrets
• Discovery ask: forensic review of devices, drives and networks (OpenAI refused this piece)
• Preservation: hold all relevant evidence
• Timeline: Apple sued July 10; injunction motion Aug. 3-4

“Apple will be irreparably harmed absent a preliminary injunction,” the company wrote. OpenAI agreed to three of the five demands (halt future access, stop use, preserve evidence) but rejected letting Apple’s lawyers and forensic experts comb its systems.

That split defines the early fight. OpenAI will accept forward looking restraints and a hold on evidence. It will not open its systems to Apple’s outside lawyers and forensic teams while the parties still contest what, if anything, was taken. Expedited depositions of Liu, Tan, Peng, the unnamed fourth former employee, and corporate representatives remain on Apple’s list as the path to test those claims under oath.

The Warning Letter That Never Reached the Right Desk

Apple’s complaint claimed it warned OpenAI in February and heard nothing. OpenAI’s post shows Weil, Gotshal & Manges partner Gabriel Gross emailed general counsel Che Chang by mistake after confusing two Asian last names, then thanked Chang for a phone call that never occurred.

Apple had claimed that they contacted OpenAI in February and that we didn’t respond. They now admit that their outside lawyers emailed the wrong person after confusing two Asian last names-only after we brought this to their attention.

Chang forwarded the note to Apple’s in-house lawyers saying he had never spoken to Gross. Gross later apologized and said the intended recipient, a former Apple employee surnamed Wang, had offered to cooperate. Apple then went silent for five months until the July suit, according to the published chain.

1. Feb. 23, 2026: Gross sends warning letter to wrong OpenAI executive
2. Same day: Accidental follow-up claiming a call that never happened
3. Immediate: Chang escalates the mix-up to Apple counsel
4. Apology: Gross notes Wang’s cooperation offer; Apple says issues are “resolving”
5. July 10: Apple files the trade-secrets complaint
6. Aug. 3-4: Injunction motion and OpenAI’s public rebuttal

OpenAI says the specific allegations in the lawsuit were never raised in those February exchanges. The gap matters for the irreparable harm story Apple now tells the court. A warning that landed on the wrong desk, followed by months without a corrected notice, undercuts the picture of a company that flagged danger and met only silence. OpenAI is using that chain to argue that the suit’s core claims arrived for the first time in July, not in a ignored February alert.

Liu’s Laptop and the Colleagues Who Kept Calling

Apple alleges Liu, a former senior system electrical engineer, used a colleague’s laptop and an authentication bug to download dozens of confidential hardware files after his January 22, 2026 exit. It also claims he shared material with other Apple staff applying to OpenAI.

The iMessages OpenAI released show a different pattern. After Liu’s last day, former colleagues messaged him for help locating files, technical details and shipping decisions. One joked he was “the best” source “even if you don’t work here anymore.” Another kept his iCloud signed in to finish transfers. Liu repeatedly told them he was fine with file access but warned about iMessages from his new company appearing.

• Colleagues asked Liu to locate files and supply technical details after his exit
• Shipping decisions still routed through him in message threads
• One contact called him “the best” source even though he no longer worked there
• Another left iCloud signed in so transfers could finish
• Liu flagged the risk that iMessages from his new employer would show up

OpenAI argues any lingering access stemmed from Apple’s own offboarding, which leaves departed workers connected to systems they never asked to keep. “Former employees who are trying to do the right thing when they leave still have access to Apple files-despite not wanting them or even being aware of them,” the company wrote.

The two accounts point at different actors. Apple’s version centers on a departed engineer who sought files through a colleague’s laptop and a bug. OpenAI’s version centers on an employer that never fully cut the links, then watched remaining staff pull a former colleague back into live work. A judge will have to sort intent from residue. The messages already make the residue hard to ignore.

Tan’s Two Decades and the Parts-for-Interview Claim

Tan left after nearly a quarter-century designing core products. He co-founded io Products with Jony Ive and others; OpenAI acquired the team in a deal valued near $6.5 billion. The io Products merger letter from Sam and Jony framed it as the path to new hardware. Ive himself is not named in the suit.

Apple claims Tan circulated an internal “Need to Know” security document so new hires could skirt exit checks and that candidates brought actual batteries and logic boards to OpenAI interviews for “show and tell.” One candidate allegedly said he “didn’t even know we could take those from the office.” OpenAI says Tan has always told his team it must not touch confidential material from other companies.

Tan’s seniority sharpens both stories. A vice president who spent twenty four years on iPhone and Apple Watch design would know which supplier details matter and which exit steps catch departing staff. Apple treats that knowledge as the reason the “Need to Know” document and the parts at interviews, if proven, would be so damaging. OpenAI treats the same tenure as the reason Tan’s standing instruction against touching other companies’ material should carry weight. The $6.5 billion io path and the merger letter cast the hire as a hardware bet, not a secrets bet. Ive’s absence from the caption keeps the personal brand outside the pleadings even as the team he joined with Tan sits at the center of OpenAI’s device plans.

Four Hundred Exits and a Fragile Partnership

Apple’s complaint notes more than 400 former Apple employees now work at OpenAI, many from sensitive hardware, chip and design teams. “It is not surprising” some carry knowledge of confidential information, Apple wrote, but “that does not entitle OpenAI to use that information to jumpstart its hardware efforts.”

The same talent flow that fuels the lawsuit also supplies OpenAI’s defense narrative. Crowd reaction on X quickly zeroed in on the irony: the recruiting pipeline itself generated server logs and screenshots that became exhibits. Apple’s secrecy culture produced both the alleged secrets and the residual-access problem now on public display.

The companies still partner on ChatGPT access inside Siri and Apple Intelligence. That relationship has grown strained as OpenAI pushes for a direct consumer device that could sidestep traditional apps and the iPhone itself. Analysts have floated a phone-like AI-agent device. Apple’s suit frames the hardware push as built on looted designs, manufacturing processes and supplier techniques, including one alleged secret metal-finishing demo run under the false belief that Apple had approved it.

Four hundred exits turn a bilateral commercial partnership into a porous border. Knowledge walks with people even when files do not. Apple’s filing tries to draw a hard line between general skill and specific secrets on manufacturing and suppliers. OpenAI’s public reply tries to draw a different line between active taking and passive residual access left by the prior employer. The metal-finishing demo allegation, if developed in discovery, would test whether OpenAI’s hardware path used a process it should never have seen. Until then the headcount alone keeps pressure on both the court fight and the product partnership.

What the Injunction Would Change

A granted order could force OpenAI to prove its device work is clean component by component before any launch, delaying timelines and raising costs just as the company prepares product and potential public-market moves. OpenAI calls the request “both based on false information and completely unnecessary because we do not have, nor want, any of their trade secrets.”

The broader money and infrastructure bets continue. Nvidia’s massive OpenAI infrastructure bet and Microsoft’s parallel OpenAI hedge moves show how capital is still flooding the same ecosystem even while the two consumer brands spar in court. OpenAI earlier defeated a separate trade-secrets challenge from Elon Musk’s xAI.

Practical effects would land first on process. Clean room reviews, supplier rechecks and slowed integration with the io team would follow any order that freezes use of disputed material. Capital from Nvidia and Microsoft does not pause for a discovery fight, but a launch calendar can. OpenAI’s win against the earlier xAI trade secrets claim gives it a recent template for defending talent driven suits. Apple’s motion aims to make this case different by tying named engineers, specific file paths and a hardware roadmap together before a product reaches buyers.

The Device Race Compresses Every Deadline

Both companies are chasing the next consumer AI interface at the same time the injunction papers land. Apple’s motion treats OpenAI’s hardware work as a threat that cannot wait for ordinary discovery schedules. OpenAI’s blog treats the same work as a clean effort now slowed by claims built on a misdirected letter and leftover system access.

A phone-like AI-agent device, the kind analysts have already floated, would sit outside the app model that still routes much of ChatGPT’s reach through other companies’ platforms. That prospect turns ordinary hiring friction into a strategic filing. If OpenAI can ship without the court imposing component level proof burdens, it keeps calendar control. If Apple wins a broad freeze, the proof burden itself becomes a brake.

• Apple wants restraints in place before device work hardens into a launch
• OpenAI wants the record to show it neither holds nor wants Apple secrets
• The Siri and Apple Intelligence partnership continues under that strain
• Infrastructure backers keep funding the wider ecosystem around both firms

The Northern District timeline now runs in parallel with product timelines neither side fully controls. Each hearing date and each discovery fight becomes another input into when, and on what terms, a direct consumer device could appear.

Offboarding Gaps Feed the Public Case

OpenAI’s strongest public exhibits are not rival design files. They are Apple’s own residual laptop access, unfinished offboarding, a counsel mix-up over surnames, and five months of silence after the February misfire. Those points do not by themselves defeat a misappropriation claim. They do shape how the industry reads the suit while the judge still has only paper to go on.

Apple built its edge on obsessive control of design, supply chain and people. Control that incomplete fails in public creates a second problem: the story of the failure travels faster than the sealed exhibits. Server logs and screenshots from the recruiting pipeline became proof points because the same secrecy culture that guards hardware also leaves traces when hundreds of people leave for one rival.

Liu’s iCloud thread and the Gross-Chang email chain give OpenAI a simple narrative to repeat. Apple’s counter narrative depends on discovery that OpenAI has so far refused to allow inside its systems. Until a court orders a deeper look, the offboarding gaps remain the part of the case anyone can read without a protective order.

Process Debt Meets the Hardware Deadline

Apple built its edge on obsessive control of design, supply chain and people. The injunction fight now puts that control under a microscope. Residual laptop access, unfinished offboarding, a counsel mix-up over surnames and five months of silence have become OpenAI’s strongest public exhibits. Whether a judge finds actual misappropriation is a separate question from the optics already in the open.

The hearing calendar and full discovery will test the documents. Until then the spectacle itself does the work Apple’s warning letter never quite managed: it has the whole industry watching exactly how hard it is to keep hardware secrets when hundreds of the people who know them have already walked out the door.

Process debt and product speed now pull in opposite directions. Apple must show the court that the harm is both real and urgent. OpenAI must show that urgency rests on a faulty premise. The winner on the preliminary record will set the tone for every later fight over what the io team may build and what Apple can still keep behind the curtain.