Apple filed a trade secrets lawsuit against OpenAI on July 10, 2026, in federal court in Northern California, alleging a coordinated scheme to steal hardware IP at every level of the company. The complaint names OpenAI's chief hardware officer -- a former Apple VP of 24 years -- and involves over 400 former Apple employees now at the AI lab. Here is what the case actually says.
This is the tech story of July. I have read through the reporting and pulled the complaint details, and the allegations are more specific than the initial headlines suggested. Here is what happened, why it matters for OpenAI's trajectory, and what the case means for builders thinking about AI data law.
What did Apple actually allege happened?
Apple's lawsuit accuses OpenAI and io Products of running a coordinated program to extract hardware trade secrets -- "at every level," from technical staff to OpenAI's chief hardware officer. The alleged scheme involved directing job candidates to physically smuggle Apple hardware components into interviews and coaching departing employees on how to evade Apple's security procedures when leaving the company.
The complaint names two central figures. Tang Tan, OpenAI's chief hardware officer, spent 24 years at Apple rising to VP of iPhone and Apple Watch product design before leaving for io Products in 2024. Apple alleges Tan directed interview candidates to bring "actual parts" -- batteries, logic boards, system-in-package chips -- to their OpenAI interviews for what he called "show and tell sessions." He also allegedly used Apple's confidential project code names during the recruiting process and coached departing employees on how to bypass Apple's security teams when copying files. The second figure, Chang Liu, spent 8 years at Apple as a senior systems electrical engineer before joining OpenAI in 2026. Liu allegedly failed to return an Apple-issued laptop when he left and used it to download dozens of confidential Apple technical documents labeled as proprietary.
The complaint also reveals that OpenAI internally circulated an Apple document bearing a "Need to know" designation. That document contained specific instructions on how to avoid the "dreaded walkout" -- Apple's security process for departing employees, which involves IT confiscating devices and escorting people out of the building. Circulating that document internally is not standard recruiting behavior. That is a playbook for systematic extraction.
Over 400 former Apple employees now work at OpenAI, according to the complaint. Apple is describing OpenAI's hardware business as "rotten to its core."
Why is OpenAI recruiting this aggressively from Apple?
OpenAI is building consumer hardware and needs Apple's deep vertical expertise to do it. The company acquired Jony Ive's io Products firm for $6.5 billion in 2024 -- the same Jony Ive who designed the original iPhone, the iPod, and the MacBook Air. The ambition includes a smartphone competitor reportedly targeting a 2028 launch and a HomePod-style smart speaker. Neither product is possible without hardware engineering talent that has built consumer devices at Apple's scale before.
Apple's hardware design is notoriously hard to replicate. The supply chain relationships, the chip integration knowledge, the manufacturing tolerances -- none of it is public documentation. Apple's complaint suggests OpenAI was not just trying to hire people who happened to know this; they were allegedly trying to extract the knowledge itself as part of the recruiting process. Asking candidates to bring "actual parts" in their bags is not about evaluating technical skill. That is about getting the parts in the door.
From partnership to lawsuit: how Apple and OpenAI fell out
The two companies had a high-profile partnership announced in 2024 when ChatGPT was integrated into iOS, iPadOS, and macOS. At the time it was framed as a landmark collaboration -- Apple providing distribution at scale, OpenAI providing the intelligence layer. That partnership gave OpenAI access to hundreds of millions of Apple users overnight.
The relationship deteriorated when OpenAI announced its own hardware ambitions. The $6.5 billion io Products acquisition made clear that OpenAI was not just an AI lab anymore -- it was positioning to compete with Apple in consumer devices. From Apple's perspective, they had handed OpenAI massive iOS distribution while OpenAI was simultaneously building the hardware line to displace them. The trade secrets lawsuit is Apple's direct response to that competitive shift. The language in the complaint -- "rotten to its core," misconduct "at every level" -- signals that Apple is not interested in a settlement that lets OpenAI keep the disputed IP and continue building.
Get the AI Agent Briefing
One email per week. The best AI agent news, tutorials, and tools -- written by someone who actually builds with them.
Subscribe Free
What does this mean for OpenAI's IPO timeline?
OpenAI was reportedly targeting an IPO in 2026 -- possibly before year end. A major trade secrets lawsuit from Apple introduces legal and reputational uncertainty that directly affects underwriting timelines and investor appetite. Apple is asking the court to bar OpenAI from using or disclosing any Apple trade secrets, require return of all confidential materials, preserve all evidence related to the case, and award monetary damages.
The injunction risk is the most acute immediate problem. If Apple wins an injunction, OpenAI's hardware development program at io Products could be legally blocked from using any of the allegedly stolen IP. That is a direct threat to the 2028 product roadmap -- the entire strategic rationale for the $6.5 billion acquisition. IPO underwriters price in legal risk, and "your hardware program could be legally halted before it launches" is significant legal risk to disclose in an S-1.
What does this case mean for what builders share with AI tools?
The Apple lawsuit is primarily about physical IP theft -- hardware components smuggled out of buildings, stolen laptops, confidential files downloaded on unreturned company property. But it surfaces a related question every builder using AI tools with client or proprietary code should be able to answer: what is your AI vendor doing with the confidential information you provide, and does that information retain trade secret protection once it enters an AI system?
Courts are developing this doctrine in parallel cases. In Trinidad v. OpenAI, a plaintiff's trade secret claims were dismissed because she had voluntarily disclosed her proprietary frameworks to ChatGPT while using it. The court found she could not claim confidentiality after sharing information with a public AI platform -- treating it as disclosure to the world. The principle emerging from these cases: sharing confidential material with a public generative AI tool, without appropriate contractual safeguards, may strip trade secret protection under the Defend Trade Secrets Act.
The technical problem compounds the legal one. LLMs blend patterns from their inputs into complex internal representations -- model parameters, embeddings, attention weights. Once confidential material is absorbed into those systems, isolating and removing a single company's contribution is technically impossible without rebuilding the model. There is no "delete my trade secret" endpoint. What goes into a model's training distribution does not come back out in any recoverable form.
Three things builders should actually change
The Apple-OpenAI lawsuit will not change the daily workflow of most builders -- it is corporate IP litigation at a scale most teams do not operate at. But the underlying question is one every builder using AI tools with client code, proprietary architecture, or confidential product plans should answer before pasting anything into a chat interface.
First: enterprise API tiers with data processing agreements are legally different from public web interfaces. When you use the Claude API or OpenAI API under a DPA, those agreements contractually restrict training use of your inputs and establish data retention limits. When you paste into a browser chat on a free or personal account, you may have no such contractual protection. Most builders know this in the abstract -- few have verified which tier they are actually on before sharing sensitive work.
Second: zero data retention settings need to be verified, not assumed. Many AI platforms offer data retention controls, but these settings are often not active by default. If your work involves proprietary code or client IP, check your account settings explicitly. The difference between a setting existing in the platform and that setting being active on your specific account is the difference between protection and exposure.
Third: personal accounts and company work should not share the same sensitive information. Trinidad v. OpenAI is a direct warning -- voluntarily sharing proprietary frameworks on a personal account with a public AI tool can undermine trade secret protections entirely. Enterprise accounts with legal agreements in place change that calculation. Professional work with confidential inputs should go through your company's enterprise tier, not a personal subscription.
The Apple lawsuit is the highest-profile case yet to bring AI IP questions into mainstream legal discussion. The precedents being established here -- both in the Apple case and in parallel training data litigation -- will shape how builders and companies think about AI vendors for the next several years.
FAQ
When did Apple sue OpenAI, and where was it filed?
Apple filed its trade secrets lawsuit against OpenAI on July 10, 2026, in the Northern District of California -- the federal court in San Francisco that handles most major Silicon Valley tech litigation. The case also names io Products, the hardware design firm OpenAI acquired for $6.5 billion from former Apple design chief Jony Ive in 2024.
Who is Tang Tan and why is he central to the Apple complaint?
Tang Tan is OpenAI's chief hardware officer and a former Apple vice president who spent 24 years at Apple leading iPhone and Apple Watch product design. Apple's complaint alleges Tan directed job candidates to bring physical Apple hardware components -- batteries, logic boards, SIPs -- to their OpenAI interviews for "show and tell sessions," and coached departing Apple employees on how to evade Apple's security procedures when leaving.
Does the Apple vs. OpenAI lawsuit affect AI training data law?
Indirectly. The Apple suit focuses on physical IP theft, not training data, but it runs parallel to cases like Trinidad v. OpenAI where courts are addressing whether sharing proprietary information with a public LLM strips trade secret protection under the DTSA. Builders using AI tools with proprietary or client data should understand the legal difference between enterprise API agreements with DPAs and public web interfaces without such protections.
What remedies is Apple seeking in the lawsuit?
Apple is seeking an injunction barring OpenAI from using or disclosing any Apple trade secrets, a court order requiring OpenAI to return all confidential Apple materials, preservation of all evidence related to the case, and monetary damages. If an injunction is granted, it could directly halt hardware development at io Products -- the core rationale for OpenAI's $6.5 billion acquisition of that firm.
Get the AI Agent Briefing
One email per week. The best AI agent news, tutorials, and tools -- written by someone who actually builds with them.
Subscribe Free