Apple’s OpenAI Trade-Secret Filing: Why the Alleged AI-Agent Use Matters

Apple’s supplemental brief in its trade-secrets case against OpenAI centers on a critical alleged shift: from a former employee retaining access to Apple information to actually using a confidential circuit design while employed by OpenAI. Filed August 31 and publicly reported around September 1, the brief supports Apple’s request for expedited discovery. Its factual assertions remain allegations, not judicial findings. 1

The new allegation: post-employment use of an Apple circuit design

Apple alleges that former senior system electrical engineer Chang Liu joined OpenAI in January 2026 but retained access to Apple-linked third-party cloud storage. According to Apple, Liu downloaded a confidential power-converter circuit schematic on March 7, then ran an LTspice simulation using the schematic on March 18 under a profile named “changliu.” 1

Apple says the simulation generated output files that later synchronized through iCloud, creating artifacts on an Apple-issued MacBook. Liu’s attorneys provided that MacBook on August 21, and Apple says its initial forensic review revealed the download, simulation, and synchronization evidence. 1

That distinction is central to Apple’s litigation strategy. A claimed download alone may raise questions about access and retention; a claimed simulation after Liu began working at OpenAI is the basis for Apple’s argument that its confidential design was actively used in OpenAI-related work. 1

Why Apple highlights the AI-agent allegation

Apple also alleges that Liu used an AI agent to learn LTspice and optimize compensation parameters connected to the simulation. Reuters reported that Apple told the court Liu had accessed the circuit and trained an AI agent using proprietary Apple information in March. 1

Apple’s stated concern is that AI-assisted use can make a trade-secret dispute harder to contain. In its framing, confidential information supplied to an agent or model may influence later work or outputs in ways that are difficult to trace, retrieve, or remediate. That argument is not a court conclusion about how any particular AI system works; it is Apple’s explanation for why discovery and evidence preservation should proceed quickly. 1

The filing therefore places an AI-specific issue alongside conventional trade-secret questions: not only whether a confidential file was copied or used, but whether technical information may have been incorporated into an AI-enabled workflow in a way that complicates forensic review.

The alleged evidence-destruction issue

Apple further alleges that Liu, after learning of Apple’s June investigation, instructed OpenAI colleague Yu-Ting “Alyssa” Peng to restore Apple-issued devices and then continue using them. Apple characterizes that conduct as attempted destruction of evidence, since device restoration can remove logs, metadata, and other records that may be relevant to forensic analysis. 27

This too remains contested. The immediate legal question is not whether Apple has proved spoliation, but whether its showing justifies faster access to devices, accounts, documents, communications, and testimony before potentially relevant electronic evidence is lost.

Why expedited discovery is the immediate fight

Apple is asking Judge Edward J. Davila to authorize expedited discovery while it pursues preliminary relief. Apple argues that the recovered MacBook suggests other devices and accounts—including a Mac mini associated with the alleged simulation activity—may contain relevant evidence that has not yet been examined. 16

If the court permits accelerated discovery, Apple could seek earlier inspection and collection of relevant electronic material, along with communications and testimony needed to test its claims. The request does not decide the merits of the trade-secret case. Instead, it concerns what evidence can be preserved and examined before the ordinary discovery process runs its course.

Apple’s most consequential claim is therefore evidentiary: it says the MacBook provides a forensic connection among the alleged circuit download, LTspice work, cloud synchronization, and use after Liu joined OpenAI. Whether that connection withstands adversarial testing could influence the scope of discovery and Apple’s request for preliminary injunctive relief. 1

OpenAI’s response

OpenAI has called Apple’s trade-secret claims meritless. In its court response, OpenAI argued that Apple had not shown former employees stole confidential information and said the dispute was “a mess of Apple’s own making.”

OpenAI’s defense also points to Apple’s offboarding and information-management practices. It argues that Apple’s own handling of employee access—including its alleged use of personal accounts for work and continued access after departures—undercuts Apple’s claims.

The company has said it neither possesses nor seeks Apple trade secrets and has characterized the lawsuit as an attempt to interfere with employee recruitment and competition. Reporting indicated that OpenAI did not immediately issue a separate public comment specifically addressing Apple’s supplemental brief. 2

What happens next

The supplemental brief listed September 4 as the response deadline for defendants, with a hearing on Apple’s expedited-discovery motion set for October 1, 2026, at 9:00 a.m. before Judge Davila.

For now, the filing is significant because it gives Apple a more concrete alleged forensic narrative in support of accelerated discovery. But the key propositions—including whether Liu used Apple’s schematic at OpenAI, whether AI-agent activity materially increased the risk of misuse, and whether evidence was improperly destroyed—are disputed allegations that will need to be tested through the litigation process.