Apple’s ‘Shocking Evidence’ Against Ex-Employee Who Took Data to OpenAI

Apple has filed what it calls "shocking evidence" in its ongoing lawsuit against OpenAI, after legal counsel for former Apple engineer Chang Liu handed over Liu's old Apple work laptop earlier in August 2026. Apple now alleges Liu used a confidential circuit schematic and an internal Apple engineering tool during his work at OpenAI, and that Liu recruited OpenAI colleague Yu-Ting Peng to help destroy evidence in June when he found out Apple was investigating him. More than 400 former Apple employees now work at OpenAI, according to Apple's initial filing.

What happened

Detail	Fact
Source of new evidence	Liu’s surrendered Apple work laptop, handed over by his legal counsel in August 2026
Alleged stolen material	A confidential Apple circuit schematic and a tool sharing a name with an internal Apple engineering application
Alleged evidence destruction	Liu enlisted OpenAI colleague Yu-Ting Peng in June after learning Apple was investigating him
Former Apple staff now at OpenAI	More than 400, per Apple’s initial filing
Apple’s requested remedy	Preliminary injunction blocking OpenAI from working on Apple-derived hardware, plus expedited discovery

The laptop became the focal point after weeks of delay from the defendants, according to Apple’s court filing. The evidence itself remains redacted from public view, but Apple’s argument is clear: its trade secrets are actively in use and materials are being destroyed.

Earlier filings had already included text messages from Liu, complete with “crying laughing” emojis, showing he was aware he still had access to Apple files after leaving the company.

The access dispute: bug or bad offboarding?

The two sides tell very different stories about how Liu could still reach Apple files after leaving the company. OpenAI published a blog post earlier in August arguing that residual access after an employee departs is a common Apple problem caused by inadequate access management during offboarding. OpenAI also stated that Liu only accessed Apple files post-departure to help former colleagues who asked for his assistance.

Apple disputes this directly. According to Apple’s filing, Liu maintained access not because of sloppy offboarding, but because he “exploited a rare, previously unknown authentication bug.” That is a significant legal distinction: one framing makes Apple look negligent, the other makes Liu look actively malicious.

Why it matters

This case sits at the intersection of AI hardware development and corporate trade secret law, two areas where the stakes are extremely high. Apple is not just seeking damages. It wants a preliminary injunction, a court order that would stop OpenAI from working on any hardware product that draws on Apple’s technology for the duration of the case. That could meaningfully slow OpenAI’s hardware ambitions if granted.

The expedited discovery request is also notable. Apple argues that more former employees beyond Liu may be involved, and it wants access to evidence quickly before more is destroyed. With over 400 former Apple employees at OpenAI, the scope of any further investigation could be substantial.

For anyone watching the broader AI industry, this case signals how aggressively established hardware companies will protect their silicon and systems knowledge as AI labs push hard into custom chip and device development. The dynamics here are similar to the tensions that emerged when OpenAI’s contractor relationships shifted, but with much sharper legal teeth.

Our take

Apple’s framing of “shocking evidence” is, of course, litigation language. Courts decide what is actually shocking. But the specifics here are harder to dismiss than most corporate complaints: a named schematic, a named internal tool, a named colleague allegedly recruited to destroy evidence, and a laptop that took weeks to produce. That is a more concrete set of allegations than the usual trade secret boilerplate.

OpenAI’s “bad offboarding” defense is plausible as a general principle, but it gets weaker the moment Apple can point to a specific authentication exploit. If Apple can prove that bug in court, the narrative shifts entirely from negligence to intent.

For businesses thinking about integrating AI tools into their own operations, this case is a useful reminder that the AI products you use are built by people who came from somewhere. The provenance of training data, architectures, and engineering tooling is going to be litigated hard over the next several years. It is worth asking your vendors the uncomfortable questions now.

What to do about it

1. Review your own offboarding process: revoke access tokens, API keys, and internal tool credentials on the employee’s last day, not weeks later.
2. Audit which internal tools and schematics employees can access and export, especially senior engineers working on proprietary hardware or data pipelines.
3. If you use AI tools built by labs currently in litigation (OpenAI, others), keep an eye on injunction outcomes that could affect product availability or roadmaps.
4. Talk to legal counsel about what your IP agreements actually cover when employees move to a competitor or AI lab.

The cleanest protection against an Apple-style lawsuit on your end is airtight access management from day one, not a scramble after someone hands over a laptop.

Source: TechCrunch · AI