Apple targets OpenAI trade secrets in AI-agent evidence fight

Apple’s new target: not just files, but AI-enabled use

Apple’s trade-secret lawsuit against OpenAI has shifted from a familiar departing-employee dispute into a test of how courts may treat proprietary knowledge once it enters an AI-assisted engineering workflow. In a supplemental filing submitted on August 31, Apple told the U.S. District Court for the Northern District of California that early forensic work on a MacBook used by former Apple engineer Chang Liu uncovered evidence that he “not only downloaded” a confidential Apple circuit schematic but also used it in work at OpenAI [1]. Reuters reported the same day that Apple alleges Liu accessed a power-converter circuit schematic while at OpenAI and trained an AI agent using proprietary Apple information in March 2026 [2].

The allegations remain allegations: the court has not ruled that OpenAI, Liu or other defendants misappropriated Apple trade secrets. But the new filing matters because it gives Apple a more concrete factual theory than a broad claim that talent moved from Cupertino to an AI rival. Apple is now pointing to a specific engineering file, a specific simulation tool, a specific device trail and a specific claim about AI automation [1].

What the MacBook allegedly showed

According to Apple’s filing, Liu’s lawyers delivered the MacBook to Apple’s forensic expert on August 21, and Apple says preliminary findings began coming back on August 26 [6]. Apple claims the device contained Apple trade-secret information, including at least one “.asc” file related to a power-converter circuit schematic and simulation input data [6]. Unite.AI reported from the filing that Apple says Liu downloaded the file from Apple’s third-party cloud storage on March 7, after leaving Apple for OpenAI in January 2026 [6].

Apple says the forensic record points beyond simple possession. The company alleges that on March 18, Liu used the schematic file to run a simulation in LTspice, a tool used by electrical engineers to develop and analyze circuits [6]. Apple also says the simulation generated at least three output files uploaded to Liu’s iCloud account from a separate Mac mini, and that both the circuit file and outputs later synced to the MacBook on April 11 when Liu signed into iCloud on that device [6]. 9to5Mac reported that Apple now wants access to the Mac mini because it says the use of the schematic surfaced through iCloud syncing between the Mac mini and the MacBook [4].

That device path is central to Apple’s argument for speed. The company says defendants produced only limited information after weeks of delay and did not inspect the MacBook before advancing arguments that Apple says the device data undermines [4]. Apple is asking for expedited discovery so it can obtain documents and testimony earlier than the normal schedule, including evidence it says may show the scope of any use or destruction [2].

The AI-agent allegation

The most novel part of Apple’s claim is not that a file may have been copied; it is that Apple says the file became part of an AI-assisted workflow. Apple’s filing says messages around the same period described Liu’s AI “agent” learning how to run LTspice, look at the results and tune a compensation parameter [6]. MacRumors reported that Apple argues trade-secret information fed into an AI agent or model may create “irreversible and continually propagating” uses of the trade secret [5].

That argument is strategically important. In a conventional trade-secret case, a company may seek to recover documents, delete files, image devices and restrict employees from using specific materials. AI agents complicate that remedy if they can operationalize a workflow learned from confidential inputs. Even if the agent did not retain a full schematic, Apple’s theory suggests that the value of the secret may have been converted into repeatable process knowledge: run the simulation, inspect the output, adjust parameters and repeat.

There are still unanswered technical questions. Public reporting on the filing does not establish which AI system Liu allegedly used, whether Apple data entered a model’s long-term training set, or whether the agent retained information after the task [1]. Those distinctions matter because using an AI tool to control engineering software is different from training a general-purpose model on proprietary design data. But Apple does not need to prove all of that at the discovery stage; it needs to persuade the judge that the risk is concrete enough to justify faster evidence gathering.

Evidence preservation becomes another front

Apple’s August 31 filing also escalates the case by alleging evidence-handling misconduct. 9to5Mac reported that Apple says the MacBook’s initial forensic analysis revealed Liu sent instructions for destroying evidence to an OpenAI colleague after learning of Apple’s internal investigation, and that the colleague confirmed she would comply [4]. Bloomberg Law reported that Apple characterized the conduct as part of a broader concern that OpenAI is actively destroying crucial evidence in the trade-secret case [3].

Apple also claims Liu and others at OpenAI were aware that his access to Apple’s third-party cloud storage remained active, a point the company uses to push back against any suggestion that downloads were merely inadvertent syncing [4]. Unite.AI reported that Apple’s filing says Liu continued to access and download files through the end of April 2026, which Apple argues is inconsistent with a theory of accidental residual access [6].

OpenAI has pushed in the opposite direction. Reuters reported that OpenAI has sought to dismiss the lawsuit and says it is building products that are “entirely new” [2]. Bloomberg Law reported that OpenAI has maintained it followed industry recruiting standards and that Liu was trying to help a former Apple colleague when accessing cloud data [3]. OpenAI did not immediately respond to Reuters’ request for comment on Apple’s latest allegations [2].

Why this is bigger than one engineer

The dispute sits at the intersection of talent mobility, hardware ambition and AI tooling. Bloomberg Law reported that Liu is one of more than 400 former Apple employees who have gone to OpenAI as the AI company moves into hardware devices [3]. Apple’s concern is not simply that experienced engineers carry general know-how; every technology company benefits from lawful employee mobility. The legal question is whether specific confidential files, circuit designs, manufacturing knowledge or internal tools moved with them and were used in competing work.

That distinction will shape the case. Trade-secret law generally protects information that has independent economic value from not being generally known and that a company takes reasonable steps to keep secret. OpenAI’s defense has already questioned Apple’s security and offboarding practices, while Apple’s latest filing tries to show that the issue is not poor housekeeping but knowing use of highly confidential material [3].

For AI labs and hardware companies, the operational lesson is immediate. Access revocation, device return, cloud-token cleanup and post-employment monitoring are no longer back-office hygiene; they are litigation-critical controls. If courts accept that an AI agent can absorb or operationalize misappropriated know-how, companies may begin treating agent logs, tool-use traces, prompts, simulation outputs and local workspace files as sensitive evidence trails.

The next procedural markers are already set. Unite.AI reported that defendants may file a five-page response on September 4 and that the expedited-discovery motion is scheduled for an October 1 hearing before Judge Edward J. Davila [6]. Until then, Apple’s latest move is best understood as a bid to widen and accelerate discovery: from a MacBook to a Mac mini, from files to messages, and from copied schematics to AI systems that may have learned how to use them.

Sources from the last 72 hours

1. [1]Apple alleges OpenAI employee accessed circuit plans after joining startupAug 31, 2026, 8:41 PM UTC
2. [2]Apple Says OpenAI Is Destroying Evidence in Trade Secrets CaseAug 31, 2026, 9:03 PM UTC
3. [3]Apple reveals ‘shocking evidence’ from ex-employee’s MacBook in OpenAI suitAug 31, 2026, 8:20 PM UTC
4. [4]Apple Says Former Engineer Used Stolen Trade Secrets at OpenAI, Taught AI Agent to Run ThemAug 31, 2026, 8:22 PM UTC
5. [5]Apple Tells Court OpenAI Employee Used Confidential Circuit SchematicAug 31, 2026, 12:00 AM UTC
6. [6]Plaintiff Apple Inc.'s Supplemental Brief in Support of its Motion for Expedited DiscoveryAug 31, 2026, 12:00 AM UTC

AI-generated article based on recent web research, then preserved as a dated editorial snapshot.