OpenAI Moves to Kill Apple's Trade Secrets Lawsuit

OpenAI asked a federal judge on Wednesday to throw out Apple's trade secrets lawsuit, and the motion lands on the weakest joint in Apple's complaint: it argues Apple never actually said what the secrets are. The 31-page filing went in August 5 in the Northern District of California and spread across the wires this morning. Its core line: "OpenAI has no use, need or desire for Apple's trade secrets. OpenAI is building something entirely new."

That is the legal argument. The tone is a separate matter. OpenAI calls the case "careless, aggressive and oddly personal," and hands Apple back its own phrase, the one from the July complaint describing OpenAI's hardware business as "rotten to its core." Some form of the word "fail" appears close to 50 times in 31 pages. This is not a defense written only for a judge.

RelatedApple Tells 40 Ex-Staff at OpenAI to Preserve Data

• OpenAI says Apple "failed to sufficiently describe the information for which it claims trade secret protection," which is a pleading defect, not a factual dispute.
• The motion frames the suit as cover for Apple's own problems, arguing Apple is compensating for "shortcomings in the market for talent and retaining its employees, and its failures to integrate AI into its products."
• A judge in the same district already dismissed a near-identical trade secrets theory against OpenAI twice, most recently without leave to amend.

What is OpenAI actually arguing?

Strip the rhetoric and three arguments remain, stacked so that Apple has to win all of them.

First, identification. Under the Defend Trade Secrets Act and its California analogue, a plaintiff has to point at a specific thing: a design tolerance, a manufacturing process, a supplier arrangement, a spec. Apple's July 10 complaint describes categories, hardware designs, manufacturing processes, supply chain strategies. OpenAI's position is that categories are not secrets, and a complaint built on them cannot survive a motion to dismiss.

Second, misappropriation. Even a well-identified secret needs facts showing it was acquired, used or disclosed by improper means. OpenAI says the complaint asserts this rather than pleads it.

Third, the corporate defendants. Tang Yew Tan and Chang Liu are named individually, but the money and the injunction risk sit with OpenAI Group PBC, OpenAI Foundation and io Products. Getting from "two former employees knew things" to "the company knowingly took and used them" is a real leap, and it is the one that has already sunk a case against OpenAI in this district.

Why does "identify the secret" decide so many of these cases?

Because trade secret law refuses to protect a vibe. A patent is published and bounded. A trade secret is bounded only by what the owner says it is, so courts make the owner say it, early and precisely. California codified the instinct in Section 2019.210 of its Code of Civil Procedure, which requires a plaintiff to identify the claimed secrets with reasonable particularity before discovery starts, and federal judges in the Northern District routinely apply that logic to cases carrying California claims.

The reason is practical. Without that gate, any company could sue any competitor that hired its people, then use discovery to look through the competitor's actual engineering work and decide afterwards what to call stolen. Apple has more than 400 former employees at OpenAI by its own count, and it sent preservation letters to roughly 40 of them in July. Broad identification plus a broad evidence hold is exactly the shape courts watch for.

None of that means Apple loses. It means the fight for the next few months is about the complaint, not about whether anything was taken.

What did the xAI case already establish?

This is the part most coverage skips, and it is the strongest thing in OpenAI's favor. Elon Musk's xAI sued OpenAI on the same basic theory, that departing employees carried confidential material to a rival. Judge Rita Lin of the Northern District of California dismissed the amended complaint on February 24, 2026 with leave to try again. xAI amended. On June 15, 2026 she dismissed it again, this time without leave to amend, holding that xAI had pleaded at most passive receipt of information, which is not misappropriation under the Defend Trade Secrets Act, and that its theory about OpenAI's recruiting rested on speculation.

Apple's complaint is stronger on facts than xAI's was. It alleges Tan directed candidates still employed at Apple to bring actual parts to show and tell sessions, and that Liu kept an Apple laptop after leaving and had used it to download confidential documents. Those are specific acts by named people. The open question is whether specific acts by two individuals add up to corporate misappropriation by OpenAI, which is the precise gap Judge Lin found fatal in the xAI case.

RelatedApple Sues OpenAI Over Stolen Secrets for AI Hardware

How did this get so public so fast?

OpenAI did something unusual on August 4, the night before the motion. It published a blog post titled "Apple is getting this wrong" and attached receipts: the email chain in which Apple's outside counsel, per OpenAI, contacted the wrong person in February after mixing up two similar surnames, complete with counsel's apology, and iMessage exchanges showing Apple employees asking Liu to locate files and move technical information after his January 22 departure, with messages running as late as March 5.

If those exchanges are what OpenAI says they are, they do double duty. They undercut the picture of a departing engineer smuggling data out, and they suggest Apple's own staff kept pulling on that thread for weeks. Apple has not responded publicly to the motion.

1. Jan 22, 2026Chang Liu leaves Apple Later named as a defendant
2. Feb 2026Apple counsel emails the wrong person Per OpenAI's published chain
3. Mar 5, 2026Last iMessage OpenAI published Apple staff still requesting files
4. Jul 10, 2026Apple sues in N.D. Cal. OpenAI, io Products, Tan, Liu
5. Jul 17, 2026Preservation letters to ~40 ex-staff Evidence hold widens
6. Aug 4, 2026OpenAI publishes its rebuttal "Apple is getting this wrong"
7. Aug 5, 2026Motion to dismiss filed 31 pages
8. NextApple opposes or amends Then a hearing on the motion

What it means for the market

Very little moves on a motion to dismiss, and that is the point worth making. Neither Apple nor Microsoft, OpenAI's largest commercial partner, faces a numbers change from this filing. What the case does affect is timing risk around OpenAI's first consumer hardware product, the reason io Products was acquired in the first place. An injunction is the only outcome here that would genuinely delay a launch, and injunctions require identified secrets. Every month the complaint spends unamended pushes that risk further out.

The signal for investors is narrower than the headlines suggest: watch whether the court orders a particularity showing before discovery. If it does, Apple's leverage drops sharply, because the discovery fishing expedition is worth more to Apple than damages ever will be. Apple's own AI position is unchanged either way, and its existing Siri and ChatGPT arrangement is explicitly not part of the suit.

Our take

OpenAI wrote two documents this week. The blog post is aimed at engineers deciding whether joining OpenAI means inheriting a lawsuit, and it is effective at that. The motion is aimed at a judge, and it is doing something more specific than it looks: it is quietly asking the court to apply the reasoning that already ended xAI's case in this same district, where good facts about individuals never reached the company.

Apple's complaint has better facts than xAI's did. Show and tell with actual parts is not speculation. But better facts about two engineers still have to travel to a corporate defendant, and Apple's decision to plead in categories gave OpenAI the cleanest possible opening. Expect an amended complaint with much more specific secret identification, and expect it to be filed under seal.

Original analysis by GenZTech. Reporting on the August 5 motion via The Verge and Reuters.