Do Son 
The trade secret lawsuit between Apple and OpenAI recently escalated. According to a recent Reuters report, Apple submitted new court documents to seek expedited discovery. These files reveal shocking evidence extracted from a former employee’s laptop. The engineer allegedly used Apple trade secrets for OpenAI hardware projects. He also attempted to direct colleagues to destroy evidence during the investigation.
Exposure of Crucial Evidence
The central figure in these new allegations is Chang Liu. He is a former senior systems electrical engineer. Liu left Apple in January and joined OpenAI. Apple claims Liu retained his company-issued MacBook. His lawyer returned the laptop on August 21. A preliminary forensic analysis of this laptop revealed severe violations.
Liu downloaded and used confidential files without authorization. Apple claims Liu knew he still had access to third-party cloud storage. He downloaded numerous confidential engineering files in March. These files included power converter circuit schematics. He then simulated these circuits using LTspice software.
Training AI with Confidential Data
Liu allegedly trained AI agents using this confidential data. He boasted to a colleague in a message. He claimed his AI agent learned to run LTspice within an hour. The agent could view results and adjust compensation parameters. The analysis showed Liu duplicated internal tools. He used a software tool at OpenAI sharing the exact name of an Apple proprietary application.
Expanding the Scope of Allegations
Apple accuses Liu of conspiracy to destroy evidence. Liu allegedly learned he was the target of an internal investigation. He then sent instructions to an OpenAI colleague. He asked this person to restore any Apple-issued devices. This action aimed to erase damaging digital evidence. The colleague reportedly agreed to cooperate.
Apple emphasized this point in their filing. Liu and other OpenAI employees clearly understood his unauthorized access. Apple argues its trade secrets face a high risk of continuous use and destruction. Therefore, the company strongly demands an expedited discovery process. Apple also expanded the lawsuit scope earlier this month. They added more names to the list of former employees. Eleven people are now accused of leaking secrets to OpenAI.
OpenAI Strikes Back
OpenAI demanded the court dismiss this case entirely. The company counterattacked the allegations as baseless. OpenAI previously issued a statement criticizing the lawsuit. They called it careless, aggressive, and oddly personal. They reiterated that OpenAI does not want Apple trade secrets. The company simply aims to build innovative products.
OpenAI previously offered a different explanation for the access. They claimed Liu retained access due to severe flaws in Apple’s departure procedures. One former Apple colleague even messaged Liu after his departure. This colleague actively asked Liu for help finding internal files.
A New Legal Test for AI Training
This lawsuit appears to be a typical talent poaching dispute. It also looks like a standard hardware secret defense battle. However, it generates a completely new legal proposition. This proposition could deeply impact the entire AI industry.
Apple presents a highly destructive viewpoint in this case. Feeding trade secrets to an AI model causes irreversible harm. The learning process continually propagates the infringement.
The Irreversible Pollution Theory
Past trade secret thefts usually involved copied hard drives. Thieves directly used the data for competing designs. Now, AI models absorb this confidential logic during training. The LTspice simulation parameters represent a prime example. Developers will struggle to prove their models do not use this stolen knowledge.
The court might ultimately accept Apple’s irreversible pollution theory. If so, AI development companies will face severe compliance pressures. They will endure extreme scrutiny when hiring competitors’ employees. Filtering training data will become exceptionally difficult. A single confidential document fed to an AI could ruin everything. The entire model might face the risk of forced destruction.
Support Our Threat Intelligence
Find our zero-day alerts and CVE reports helpful? Support our work today and unlock a 100% ad-free reading experience!
Buy Me a Coffee
PayPal
