Apple Says More Ex-Employees May Have Taken Confidential Data to OpenAI

Apple has told a federal court that at least 11 additional former employees may have taken confidential data to OpenAI — on top of the two named in its original July complaint — and is now seeking a preliminary injunction to freeze OpenAI's entire device roadmap before any allegedly stolen intellectual property ships in a product. The escalation, reported by TechCrunch on August 4, transforms what began as a targeted two-defendant filing into a sprawling trade secrets case that implicates a significant slice of Silicon Valley's ongoing AI talent migration.

What's Converging

The AI hardware race has been quietly draining Apple's engineering ranks for the better part of two years. According to Apple's own court filing, more than 400 of its former employees now work at OpenAI — a number that 9to5Mac highlighted when the original complaint dropped on July 13. Apple's filing notes bluntly that with that many former insiders in place, "it is not surprising" that OpenAI personnel carry knowledge of confidential Apple technology, but argues that employment doesn't confer a license to use it.

The talent movement is not random. Bloomberg reporting cited in 9to5Mac's coverage indicates that most of OpenAI's Apple recruits came specifically from the engineering division now led by incoming Apple CEO John Ternus — the team responsible for Apple's most sensitive hardware work. The most prominent recent example is Paul Meade, who oversaw Apple Vision Pro and Apple's smart glasses project before being poached by OpenAI. That recruitment pattern — systematically targeting engineers close to unreleased hardware — is exactly what Apple's expanded complaint is designed to address legally.

This case sits inside a broader industry tension that has been building since early 2025: the collision between aggressive AI-era hiring and decades of trade-secrets law written for slower-moving industries. When an engineer moves from one company to another, what knowledge travels with them, and when does that constitute theft? Courts have long struggled with this question. What makes the Apple-OpenAI matter different is the sheer scale of the alleged pattern — and the specificity of the methods Apple says it can now document.

The Specific Development

The core of the original complaint centered on Chang Liu, a senior systems engineer, and Tang Yew Tan, who became OpenAI's Chief Hardware Officer. Apple alleged that both had taken confidential information — either before or during their departures. The August 4 filing expands that picture considerably: Apple has identified 11 more former employees it believes may have participated in similar conduct, and says the methods were not sophisticated. Employees allegedly took screenshots of confidential documents showing unannounced Apple products before sitting for OpenAI job interviews. Others retained Apple-issued work devices after their employment ended.

Perhaps the most telling signal Apple points to: in the weeks after the original July complaint became public, several former employees proactively reached out to return Apple hardware they had been holding. Apple's legal team treats that wave of voluntary returns as evidence that awareness of the alleged misconduct was not limited to Liu and Tan — that the behavior was known more widely than the original two-defendant complaint suggested.

Apple's legal ask has also sharpened. Rather than waiting for damages to accumulate, Apple is seeking a preliminary injunction that would block OpenAI from shipping any device or product that relies on the allegedly stolen IP. That target includes not just OpenAI itself but also io, the hardware startup founded by designer Jony Ive that is now tightly linked to OpenAI's device ambitions. Apple named io as a subject of expedited discovery, suggesting it believes the alleged information may have flowed there as well.

OpenAI's public response pushes back on every element. The company denies possessing any Apple trade secrets, disputes the procedural details of how Apple served its complaint (claiming Apple contacted the wrong person and misrepresented who its general counsel was), and most pointedly argues that the "residual access" problem at the center of the exfiltration claims was Apple's own security failure — not something OpenAI engineered or exploited. Our read is that this counter-framing is strategically significant: if OpenAI can establish that Apple's own offboarding procedures left credentials and device access open for departing employees, it shifts part of the liability narrative back onto Apple and potentially weakens the injunction argument.

What's Likely Next

The immediate question is whether Apple's preliminary injunction request succeeds. To win at that stage, Apple must show it is likely to succeed on the merits and that it will suffer irreparable harm if OpenAI ships a device in the meantime — a high bar, but one Apple has now built considerable factual groundwork to support. The 11 additional witnesses and the post-complaint device returns give Apple a materially stronger record than it had in July. OpenAI's response brief and any depositions that follow will be the first real test of how durable that record is under adversarial pressure.

Separately, the io thread is worth watching closely. Jony Ive's hardware startup occupies an unusual position: it is not a subsidiary of OpenAI but is building devices intended to work with OpenAI's models, and it employs people from overlapping talent pools. If expedited discovery produces documents showing that Apple-specific design or engineering knowledge reached io before or during its product development, the case could expand again — and the injunction could reach further than OpenAI's current product roadmap. The 30-to-90-day window here is likely to produce depositions, a judge's ruling on the injunction motion, and possibly another round of complaint amendments as Apple continues to work through the devices its former employees may have held onto.

Sources