Apple’s Forensic Evidence on a MacBook Shows Liu Trained AI on Stolen Schematics

Apple has filed new forensic evidence in its trade secret lawsuit against OpenAI, alleging that former senior electrical engineer Chang Liu downloaded confidential circuit schematics, used them in his work at the AI company, and destroyed evidence after learning of Apple’s internal investigation into his conduct.

The filing, disclosed August 31, centers on forensic analysis of a MacBook that Liu took from Apple after departing in January 2026. According to Apple’s court documents, the laptop shows that Liu not only accessed Apple’s confidential files through an exploited security vulnerability but also synced those files via iCloud to a Mac mini he used at OpenAI, where he ran simulations of the schematics using LTspice, an electrical engineering tool.

Apple vs OpenAI

What The Forensic Evidence Shows

Apple characterizes the MacBook analysis as revealing several layers of misconduct. In addition to the unauthorized download of the circuit schematic, Apple alleges that Liu and others at OpenAI were aware the access was unauthorized. The filing includes messages in which Liu, upon learning of Apple’s internal investigation, sent instructions to an OpenAI colleague to destroy evidence, which the colleague confirmed she would follow.

The evidence also includes Liu’s statement that an AI “agent” he developed had learned to run LTspice and review simulation results. Apple argues this represents a fundamentally novel form of intellectual property misappropriation: feeding trade secrets into machine learning models that then embed that knowledge in a way that cannot be easily reversed or extracted.

  • Downloaded file: Confidential Apple circuit schematic
  • Execution: Liu ran simulations in March using the schematic in LTspice
  • Evidence tampering: Liu sent destruction instructions to a colleague after learning of Apple’s investigation
  • AI training: Liu’s AI agent was trained to understand and execute LTspice operations using the schematics

OpenAI’s Counterargument Centers On Apple’s Security Failures

OpenAI responded on September 1 by reframing the case not as theft but as a consequence of Apple’s own negligent security practices. The company argues that Apple encourages employees to use personal iCloud accounts for work documents, fails to properly offboard departing staff, and contacts former employees for assistance after they’ve left, inadvertently giving them reasons to access company systems.

Apple may not like those choices, but it cannot claim those choices are unlawful, and it cannot use its own sloppy procedures to blame others for its own mess.

OpenAI’s legal strategy hinges on a threshold question in trade secret law: whether Apple took adequate measures to protect the information in question. Under the Uniform Trade Secrets Act, information qualifies as a trade secret only if the owner has taken reasonable steps to maintain its secrecy. If OpenAI can establish that Apple’s access controls were insufficient, the company could prevail even if use of the information is proven.

The company does not appear to deny that confidential Apple information reached OpenAI, but rather argues that Apple’s own procedural failures, not OpenAI’s misuse, are the root cause.

The Liu case sits within a larger competitive reality. OpenAI has disclosed in court filings that it has hired approximately 400 former Apple employees for its hardware projects. Chang Liu worked at Apple for eight years before his departure. A second departing executive, Tang Tan, served as VP of product design at Apple, where he led iPhone and Apple Watch design, before leaving in February 2024 to join Jony Ive’s design venture.

Apple’s argument that trade secrets fed into AI models create “irreversible and continually propagating uses” establishes a precedent concern for the broader industry. If Apple prevails, it will have demonstrated that AI companies cannot safely hire from competitors without risk of liability, even if no deliberate theft occurs.

What Happens Next

Judge Edward J, davila is scheduled to hear arguments on October 1, 2026. At that hearing, the court will consider Apple’s motion for a preliminary injunction as well as OpenAI’s motions to dismiss the lawsuit. A granted injunction could effectively freeze OpenAI’s hardware development pending trial; a denial would signal judicial skepticism of Apple’s claims or sympathy for OpenAI’s defense regarding Apple’s security posture.

The October 1 hearing represents the case’s first critical juncture. What the judge decides about expedited discovery and preliminary relief will shape both the timeline and trajectory of a lawsuit that, win or lose, will establish new norms around hiring from competitors and the legal status of trade secrets embedded in AI models.

Newsletter
Never miss an Apple story
One email a day, the news that matters. No spam, unsubscribe anytime.
About the Author

Imran Hussain is the founder and editor of iThinkDifferent, which he launched in 2008 to cover Apple news, reviews, and how-to guides. He has spent over 15 years writing about iOS, macOS, and the wider Apple ecosystem, with a focus on hands-on guides - installing developer betas, troubleshooting, and walking through new features on his own devices. Based in Dubai, he also loves to cover photography, gaming, and the tech industry more broadly on his social media profiles.

1 comment

Leave a Reply