Apple Sues OpenAI Over Hardware Trade Secrets, Turning an AI Partnership Into a Legal Fight

Apple sues OpenAI over hardware trade secrets in a 41-page complaint containing allegations that extend far beyond ordinary employee departures. Apple claims former staff downloaded confidential files, retained company equipment, coached recruits around security checks, and brought physical components into OpenAI interviews.

The filing transforms an uneasy partnership into a direct fight over the future of consumer computing. Apple still distributes ChatGPT through its operating systems, yet OpenAI is developing hardware that could compete for the same users and daily interactions.

OpenAI denies wanting another company’s secrets. None of Apple’s allegations has been tested in court, and the defendants have not yet presented their complete factual response.

That distinction matters. Apple has supplied a detailed narrative, but a complaint presents one side of a dispute. The case now turns on whether Apple can connect specific protected information to OpenAI’s hardware program, rather than merely showing that former employees carried valuable experience into new jobs.

The dispute also puts Jony Ive’s hardware project under a harsher spotlight. OpenAI acquired io Products, the company co-founded by Ive and several Apple veterans, as part of a deal valued near $6.5 billion.

What looked like OpenAI’s boldest move beyond software now carries a legal question: did it build its hardware program by recruiting expertise, or by acquiring confidential Apple materials?

What Apple Accuses OpenAI and Former Employees of Taking

Apple’s strongest allegation is not simply that OpenAI hired its engineers. It is that OpenAI allegedly created a repeatable process for extracting protected information.

Apple filed the case on July 10, 2026, in the Northern District of California’s San Jose Division. The defendants include OpenAI, io Products, former Apple engineer Chang Liu, and OpenAI chief hardware officer Tang Yew Tan.

The federal complaint asserts claims under the Defend Trade Secrets Act and California law. Apple also accuses Liu and Tan of breaching intellectual property agreements signed during their employment.

Liu spent more than eight years at Apple as a senior systems electrical engineer. He worked on sensitive iPhone development programs before joining OpenAI in January 2026, according to the filing.

Apple alleges Liu failed to return an authenticated company laptop after leaving. It also claims he accessed another employee’s Apple computer and exploited an authentication bug to enter shared network folders.

The company says Liu downloaded dozens of confidential hardware files over several weeks. Those materials allegedly covered unreleased products, engineering presentations, technical specifications, circuit-board manufacturing, testing methods, and proprietary project data.

Apple says it found messages in which Liu celebrated discovering continued access. It also alleges he told an Apple employee being recruited by OpenAI how to avoid problems with the security team.

According to Apple, Liu directed that employee toward confidential materials concerning an unannounced product before an OpenAI interview. The filing says he moved some communications to a separate messaging application.

The allegations against Tan reach further into OpenAI’s recruiting and supplier operations. Tan worked at Apple for 24 years and most recently served as vice president of product design for the iPhone and Apple Watch.

Apple claims Tan emailed himself supplier information before leaving. It says he later used an internal project codename when questioning an Apple employee about an unreleased product.

The most striking claim involves interview “show and tell” sessions. Apple alleges candidates were instructed to bring actual components, prototypes, and computer-aided design materials into interviews.

The components allegedly included batteries, logic boards, and shields. One candidate reportedly expressed surprise that employees could remove such parts from Apple’s offices.

OpenAI interviewers also allegedly asked candidates about component selection, system integration, simulation tools, suppliers, and vendor communications. These topics can include general engineering knowledge, but Apple argues the questions targeted confidential details.

Apple further claims OpenAI personnel circulated an internal document describing Apple’s departure-security procedures. The filing alleges new hires used that information to delay disclosure of their destination and preserve access to internal systems.

The complaint extends to Apple’s manufacturing partners. Apple alleges io caused one supplier to perform a confidential metal-finishing technique after creating the impression that Apple had authorized the work.

Another supplier allegedly received narrowly targeted questions using internal terminology related to batteries and power systems. Apple argues outsiders would not know to ask those questions without confidential information.

These remain allegations. However, their specificity creates a more serious dispute than a general complaint about aggressive hiring. Apple has identified people, messages, devices, file categories, interview practices, and supplier interactions that can be examined during discovery.

Apple Sues OpenAI Over Hardware Trade Secrets as Their Partnership Fractures

The lawsuit exposes a relationship that shifted from AI partnership to platform rivalry in roughly two years.

Apple and OpenAI announced their consumer partnership in June 2024. Under the ChatGPT integration, Siri could ask users for permission before sending certain requests to OpenAI.

That arrangement gave Apple access to a prominent generative AI service without requiring every request to run through Apple’s own models. It also gave OpenAI distribution across Apple’s installed hardware base.

The commercial logic remains understandable. Apple controls devices, operating systems, and customer relationships. OpenAI controls a widely used AI assistant and the models behind it.

Yet the same arrangement contains an obvious strategic conflict. Every ChatGPT request originating from Siri reminds users that Apple relies on an outside AI provider for some answers.

OpenAI’s hardware ambitions deepen that conflict. The company no longer wants to exist only as a service inside devices controlled by Apple, Microsoft, Google, or other platform owners.

A dedicated OpenAI device would provide direct access to users, sensors, contextual data, and recurring interactions. It would also let OpenAI design an interface around its models instead of fitting them into existing software conventions.

Apple’s business depends on controlling that integrated experience. Its advantage comes from combining industrial design, custom chips, operating systems, services, retail, manufacturing, and global distribution.

OpenAI’s move into consumer hardware therefore threatens more than a single product category. It challenges the position of the smartphone as the main gateway to digital services.

OpenAI acquired io Products after working with Jony Ive on a new family of AI devices. The transaction was valued near $6.5 billion, according to Associated Press reporting.

Ive is not named as a defendant in Apple’s complaint. That omission is important because the case targets alleged conduct by Tan, Liu, OpenAI, and io, not every former Apple employee involved.

OpenAI has recruited more than 400 former Apple employees, according to Apple’s filing. That number signals the scale of the talent transfer, but it does not independently prove misconduct.

California generally supports employee mobility. Engineers can apply experience, judgment, and skills acquired during earlier jobs. A company does not own everything an employee knows.

Trade secret law draws a narrower boundary. It can protect specific information that derives value from remaining secret and that a company takes reasonable measures to secure.

Apple must therefore distinguish its protected manufacturing information from general hardware expertise. OpenAI can challenge whether the alleged information qualifies as a trade secret, whether it was used, and whether corporate leadership authorized any misuse.

The pressure on OpenAI is immediate even before those questions are resolved. A hardware team preparing products must now preserve communications, document independent development, and reassess supplier relationships.

Engineers may need to establish which design decisions came from public knowledge, fresh experimentation, or legitimate experience. That process can slow a project even without an injunction.

Apple also faces pressure. Its lawsuit draws attention to the number of experienced employees who left for OpenAI and to the possibility that a partner is building a competing device.

The case therefore serves two purposes. Apple is seeking legal protection for alleged secrets, while publicly warning suppliers and employees that its hardware knowledge cannot travel freely.

The Real Fight Is Over OpenAI’s Shortcut From Models to Manufacturing

OpenAI can train an advanced model without knowing how to manufacture millions of reliable consumer devices. That operational gap sits at the center of Apple’s case.

Software companies can ship updates quickly, observe failures, and deploy corrections through the cloud. Consumer hardware requires decisions that become expensive or irreversible much earlier.

Battery chemistry affects weight, thermal performance, charging, safety, and enclosure design. A logic-board choice can reshape component placement, radio behavior, manufacturing yield, and repair options.

A metal finish that appears cosmetic may depend on tightly controlled materials, equipment, tolerances, and supplier processes. Those details determine whether a prototype can become a repeatable product.

Manufacturing yield measures the share of produced units that meet required specifications. Low yield can make an otherwise impressive design unsuitable for commercial production.

Apple has spent decades coordinating these tradeoffs across engineering teams and suppliers. Its complaint argues that OpenAI tried to compress that learning curve by obtaining confidential materials and insider access.

That claim explains why Apple emphasizes manufacturing presentations, supplier identities, testing methods, physical parts, and internal terminology. The alleged secrets concern execution, not merely the visual form of a future device.

OpenAI’s challenge is substantial even if every allegation fails. Earlier AI devices have shown that access to a capable model does not automatically produce a compelling consumer product.

Humane’s AI Pin struggled with performance, heat, battery life, and an unclear advantage over smartphones. Rabbit’s R1 attracted attention but faced questions about reliability, software depth, and practical value.

Those examples do not determine OpenAI’s outcome. OpenAI has greater model access, capital, distribution potential, and brand recognition than most hardware startups.

Jony Ive’s design organization also brings considerable product experience. OpenAI’s io team includes veterans who understand how consumer devices move from concept to production.

However, hiring experienced people is not equivalent to inheriting Apple’s confidential documentation. This distinction will shape both the litigation and public interpretation of OpenAI’s program.

A former engineer can remember that certain designs failed or that specific manufacturing constraints matter. Apple cannot automatically convert all professional memory into company property.

The legal risk grows when that engineer downloads files, retains devices, brings components to interviews, or uses internal documents to bypass security. Apple alleges conduct across each category.

OpenAI must show that its device concepts, engineering decisions, and supplier processes were independently developed. Records created before disputed information arrived could become especially important.

Clean-room development is one possible safeguard. It separates teams and restricts access so a company can demonstrate that contested information did not influence a product.

OpenAI has not publicly described whether it used such controls. Its initial statement addressed intent, not the detailed safeguards governing former Apple employees.

The company may argue that Apple’s descriptions are overbroad or that the information never entered relevant product work. It may also challenge Apple’s attempt to attribute individual conduct to the entire organization.

Apple must establish that connection. Showing an employee improperly accessed files can support claims against that employee, but corporate liability requires additional factual and legal links.

The complaint tries to build those links through allegations involving Tan, interview practices, suppliers, and institutional departure guidance. Apple portrays the behavior as coordinated from technical staff through hardware leadership.

That institutional theory raises the stakes. If a court accepts it, an order could affect more than a handful of files. It could require broader restrictions across OpenAI’s hardware operations.

The historical comparison is Waymo’s trade secret case against Uber. That lawsuit followed Uber’s hiring of engineer Anthony Levandowski and allegations involving thousands of downloaded files.

Uber eventually settled during trial and provided Waymo with an equity payment then valued at roughly $245 million. The case showed how disputed technical materials can disrupt a heavily funded hardware program.

Apple’s allegations differ, and the outcome is not predetermined. Still, the comparison illustrates why provenance matters when a new hardware team recruits aggressively from an established competitor.

Apple’s Detailed Complaint Still Leaves Major Gaps

A vivid complaint can shape the public narrative, but Apple must prove secrecy, acquisition, use, and harm with evidence tested by the defendants.

OpenAI’s public response has been brief. Spokesperson Drew Pusateri said the company has “no interest in other companies’ trade secrets” and remains focused on building technology for users.

That statement rejects Apple’s framing but does not answer individual allegations. It does not address the retained laptop, network access, interview components, supplier contacts, or departure document.

OpenAI may provide fuller answers in court. Until then, readers should avoid treating Apple’s factual claims as established findings.

The first uncertainty concerns what the disputed files contained. Apple understandably avoids publishing its alleged secrets in an open complaint, but that limits outside evaluation.

Courts often use sealed filings and protective orders for sensitive evidence. Apple will still need to identify its trade secrets with enough precision for OpenAI to defend itself.

Descriptions such as manufacturing processes or supplier information can cover many materials. Some may be secret and valuable, while other details may be known across the industry.

The second gap concerns use. Accessing confidential files does not necessarily prove that specific information entered a commercial design, although unauthorized access can create separate contractual or legal exposure.

Apple alleges use through supplier interactions, interview questions, and OpenAI’s development work. Discovery will test whether internal documents and technical decisions support that narrative.

The third issue is corporate attribution. Apple’s complaint treats alleged acts by individuals and io as part of an OpenAI-wide strategy.

OpenAI can dispute that interpretation. It may argue that any improper conduct was unauthorized, isolated, irrelevant to product development, or addressed after discovery.

Internal policies, training materials, access logs, and management communications will matter. So will evidence showing when OpenAI learned about Apple’s concerns and what it did afterward.

Apple says it contacted OpenAI in February 2026 and received no response. If accurate, that timeline could undermine an argument that OpenAI acted quickly after receiving notice.

OpenAI may contest the nature, destination, or handling of that communication. It may also show investigative or containment steps that Apple did not know about when filing.

The fourth uncertainty involves Apple’s requested remedy. Apple seeks damages, return of property, and an injunction preventing further possession or use of its secrets.

An injunction could force OpenAI to quarantine information, change suppliers, remove staff from projects, or redesign parts of its development process. The exact scope would depend on the evidence and judicial findings.

Courts do not normally block lawful competition merely because employees changed jobs. Any remedy must connect to protected information and threatened or actual misappropriation.

This is where Apple’s reputation for aggressive litigation requires context. Apple has used copyright, patent, contract, and trade secret cases to protect products across several decades.

Its smartphone patent battle with Samsung became one of the industry’s defining legal conflicts. Earlier litigation against Microsoft challenged elements of graphical user interfaces.

Those histories show that Apple treats intellectual property as a strategic asset. They do not prove that its current claims are either valid or excessive.

The lawsuit can also function as competitive pressure even if Apple wins only a narrow remedy. Litigation consumes leadership attention, delays decisions, and makes suppliers more cautious.

It may discourage current Apple employees from joining OpenAI’s hardware group. It may also make OpenAI recruiters more conservative when discussing previous work with candidates.

For workers, the case highlights a difficult boundary. Employees should preserve their transferable skills while leaving behind documents, prototypes, credentials, source files, supplier records, and internal communications.

Companies also need better systems for separating institutional knowledge from protected materials. A searchable technical knowledge base can clarify ownership and access, but controls must follow employees through departure.

The central skeptical conclusion remains simple. Apple has presented unusually detailed accusations, yet the public record still lacks OpenAI’s complete defense and the evidence needed to test causation.

Three Signals Will Show Whether the Case Delays OpenAI Hardware

The next phase will be decided by court filings, product timing, and supplier behavior, not by the sharpest sentence in Apple’s complaint.

The first signal is OpenAI’s formal response. Its answer or dismissal motion should reveal which factual allegations it denies and which legal theories it attacks.

A broad denial backed by development records would weaken Apple’s claim of institutional misconduct. A narrower defense focused on legal definitions could leave more factual uncertainty.

Watch whether OpenAI argues that Apple failed to identify its secrets precisely. Also watch whether it separates Liu’s alleged actions from Tan, io, and OpenAI’s corporate leadership.

Counterclaims are another possibility, though none should be assumed. OpenAI could argue that Apple is using trade secret law to restrict lawful employee movement or delay a competitor.

The second signal is the court’s treatment of injunctive relief. Apple’s requested restrictions matter more to the product schedule than any early exchange of statements.

If Apple seeks a preliminary injunction, it would need to persuade the court that it faces likely irreparable harm before a full trial. OpenAI would emphasize the costs of disrupting independent development.

A targeted order covering named files or supplier processes would constrain OpenAI without freezing the entire program. A broader order could force significant project separation or redesign.

No immediate injunction would not mean Apple lost the case. It would mean OpenAI could continue development while discovery and legal arguments proceed.

The third signal is OpenAI’s hardware schedule. The company has discussed bringing consumer hardware to market, but its exact product design remains largely undisclosed.

Any postponed launch, supplier change, engineering reorganization, or revised product scope would invite questions about the litigation. Those changes would not automatically prove Apple’s allegations caused them.

Conversely, a product launch that proceeds with documented independent development would strengthen OpenAI’s position. It would also test whether consumers want a dedicated AI device beside their phones.

OpenAI’s hardware challenge is not only technical. The company must define a reason for carrying another device, trusting its sensors, and granting it persistent access to personal context.

Apple will continue controlling the iPhone, its operating system permissions, and many of the interfaces through which consumers reach AI. OpenAI needs a device valuable enough to loosen that control.

This legal fight complicates that effort because manufacturing depends on long relationships and quiet coordination. Suppliers may hesitate to support processes that resemble disputed Apple work.

The case may also push other technology companies to tighten recruiting procedures. Hardware interviews often require candidates to explain difficult past projects, creating a risk that useful evaluation crosses into confidential detail.

Responsible employers can set explicit boundaries. Candidates should describe their roles, reasoning, and general lessons without sharing unreleased designs, internal codenames, supplier identities, or proprietary artifacts.

Interviewers should never request components or documents belonging to another employer. Incoming employees should receive written instructions to return equipment and avoid transferring files.

OpenAI’s response to these governance questions will influence the case beyond the courtroom. A credible hardware organization must show that it can absorb experienced recruits without absorbing former employers’ secrets.

Apple faces its own test. It must show that the lawsuit protects identifiable assets rather than placing an invisible fence around hundreds of former employees.

For developers, enterprise buyers, and AI users, the outcome matters because hardware competition will determine where assistants live and what information they can access. It will shape whether AI remains an application, becomes a new device, or operates across both.

Apple sues OpenAI over hardware trade secrets at the exact moment OpenAI is trying to prove it can become a consumer hardware company. The timing makes the complaint commercially consequential even before a judge decides anything.

The question to follow is not whether Apple wrote a persuasive opening narrative. It is whether discovery shows that OpenAI built its hardware independently, and whether the device can still arrive on schedule.