OpenAI's Safety Shake-Up and Apple Lawsuit Deepen

Two threads that matter to anyone watching OpenAI’s IPO readiness intensified Saturday: a senior safety departure folded governance reporting lines into research, and Apple’s lawsuit added specific, damaging allegations against named OpenAI executives.

Key points

• Head of safety Johannes Heidecke will leave July 24, making him the sixth senior safety leader to exit OpenAI in two years. OpenAI is folding its safety teams under research VP Mia Glaese, eliminating the independent safety reporting line.
• Apple’s federal lawsuit in the Northern District of California names OpenAI Chief Hardware Officer Tang Yew Tan and former electrical engineer Chang Liu as individual defendants, alongside OpenAI and its io Products subsidiary.
• Chang Liu allegedly exploited a lingering network authentication flaw after leaving Apple, using a retained company laptop to download over 1,000 pages of confidential hardware files, according to the complaint.
• Tang Yew Tan, who spent 24 years at Apple as VP of product design, allegedly used internal Apple project names during recruiting and asked job candidates to bring physical prototypes to interviews.
• Apple is seeking injunctions, the return of all allegedly stolen property, royalties, and exemplary damages. OpenAI denies the allegations entirely.

What the safety restructuring signals for IPO readiness

The departure of Heidecke is not an isolated personnel move. OpenAI is absorbing its safety function into the research organization headed by Mia Glaese, which means safety no longer has a standalone chain of command. For prospective public investors, that structure matters. Independent safety oversight has been a key argument the company has made to regulators and the public about responsible development. Eliminating that independence, precisely as the company advances toward an IPO, hands critics a concrete governance concern rather than an abstract one.

Six safety leadership exits in two years is a pattern, not a streak of bad luck. The names and circumstances vary, but the direction of travel is consistent: the people charged with pumping the brakes have not stayed. Whether investors will price that risk into a future offering remains to be seen, but the optics heading into roadshow season are difficult.

How serious are the Apple allegations?

The 41-page complaint brings four claims under the Defend Trade Secrets Act and two breach-of-contract claims. The specificity is notable. Most trade secret suits rely on broad allegations of misappropriation. Apple’s complaint includes internal communications in which Liu allegedly messaged a former colleague to describe his unauthorized access, and it details the types of files retrieved: technical presentations and proprietary testing methods covering hardware developed through what Apple calls hundreds of billions of dollars in research investment.

The io Products angle complicates things further. OpenAI acquired io Products for nearly $6.5 billion to build consumer hardware around ChatGPT. If the complaint’s timeline holds up, OpenAI was allegedly soliciting Apple’s hardware secrets as it was simultaneously partnering with Apple to integrate ChatGPT into Siri and Apple Intelligence. That partnership, struck in 2024, is not being challenged in the lawsuit, but its existence alongside the alleged misconduct is the kind of detail that will dominate depositions.

OpenAI spokesperson Drew Pusateri said the company has no interest in competitors’ proprietary information and is focused on independent innovation. That denial is unambiguous, and the facts will ultimately be tested in court.

Musk adds noise, protesters add pressure

Elon Musk weighed in on X, accusing Sam Altman of “scamming” people and claiming OpenAI “stole all of Apple’s phone technology.” The post follows a long-running public feud between the two and adds little factual weight to the litigation, but it amplifies the story’s reach considerably.

Separately, what organizers are calling the largest AI protest in American history drew roughly 350 people to OpenAI’s San Francisco headquarters before marching to Google’s offices. The demonstration is modest by historical protest standards, but the framing as a record for AI-specific activism reflects a shift in public organizing around the technology. For a company preparing for a public listing, sustained street-level opposition is a reputational variable worth tracking, even if it carries no immediate legal or financial consequence.