Apple OpenAI Legal Battle Escalates as About 40 Former Employees Receive Legal Letters

Apple has reportedly sent legal preservation letters to about 40 former employees now working at OpenAI, sharply expanding its fight over alleged trade-secret theft. The Apple OpenAI legal battle began with claims against OpenAI and two former Apple employees. It now reaches a much larger group connected to OpenAI’s hardware operation.

The letters reportedly direct recipients to preserve documents, notes, and electronic communications that could become evidence. Some recipients were also asked to meet Apple’s lawyers, according to accounts citing the Financial Times. A preservation notice does not accuse its recipient of wrongdoing. However, its scope shows that Apple is investigating conduct beyond the two people named in its original complaint.

That expansion creates the central tension. Apple and OpenAI remain partners through ChatGPT’s integration with Apple products, yet they are becoming direct rivals in consumer hardware. Apple says more than 400 former employees now work at OpenAI. OpenAI, meanwhile, is developing devices with talent assembled around former Apple design chief Jony Ive.

The Apple OpenAI Legal Battle Has Expanded Beyond Two Defendants

Apple is turning a focused complaint into a broader evidence-preservation campaign across OpenAI’s hardware workforce.

Apple filed its lawsuit on July 10, 2026, in the US District Court for the Northern District of California. The complaint names OpenAI, io Products, former Apple engineer Chang Liu, and OpenAI hardware chief Tang Tan among the defendants.

One week later, reports said Apple had sent preservation notices to approximately 40 former employees. Those workers represent about one-tenth of the more than 400 former Apple employees whom Apple says now work at OpenAI.

The legal letters matter because preservation is an early foundation of civil discovery. Once a person reasonably anticipates litigation, relevant records should not be deleted, altered, or allowed to disappear through routine retention policies. The requested material can include email, messages, meeting notes, files, device records, and communications with recruiters or colleagues.

According to the report on the legal letters, recipients were told to retain potentially relevant documents and communications. The action reflects Apple’s belief that the alleged movement of confidential information extended beyond the original defendants.

That is still an allegation, not an established fact. Receiving a preservation letter does not mean someone took documents, disclosed secrets, or participated in a coordinated scheme. Companies often send such notices broadly because they do not yet know which witnesses possess relevant information.

The scale nevertheless changes the case. Apple is no longer signaling that it views the dispute as an isolated departure involving two employees. Its lawyers appear to be mapping hiring, interviews, communications, technical work, and relationships among a wider group.

The original complaint makes sweeping claims about that network. Apple alleges OpenAI systematically pursued confidential information while recruiting people from Apple’s hardware organization. It claims the conduct reached different levels of OpenAI and supported an effort to accelerate consumer-device development.

Apple is seeking orders that would prevent the defendants from possessing, using, or disclosing its trade secrets. It also wants Apple materials preserved and returned, alongside damages tied to alleged misappropriation and contractual breaches.

Those remedies explain why preservation is so important. To obtain a meaningful injunction, Apple must identify protected information with sufficient precision. It must also connect that information to defendants’ conduct and show a legally recognized threat of use or disclosure.

The approximately 40 recipients could hold evidence supporting Apple’s theory. Their records could also undermine it by showing ordinary recruiting, independent engineering, or careful separation from Apple information. Preservation protects both possibilities.

This distinction matters because employee mobility is not itself illegal. Engineers can carry general knowledge, professional judgment, and experience to a new employer. They cannot take protected documents or disclose qualifying trade secrets. The difficult legal work lies in separating those categories.

Apple’s wider campaign therefore creates pressure on both companies. OpenAI faces deeper scrutiny of how it recruited and organized its hardware team. Apple must convert its dramatic narrative into specific evidence that survives judicial review.

Why OpenAI’s Hardware Program Is Now Under Pressure

The most serious risk for OpenAI is not the letters themselves, but the possibility that disputed Apple information touched its device-development process.

OpenAI has spent years moving beyond a business centered on software models and developer services. Consumer hardware offers another route to reach users, collect contextual signals, and shape how people interact with artificial intelligence.

That strategy gained credibility when OpenAI acquired io, the hardware company associated with Jony Ive, in 2025. The transaction brought a prominent design organization and former Apple talent closer to OpenAI’s product ambitions.

Ive is not named as an individual defendant in Apple’s complaint. However, io Products is part of the litigation, and several former Apple veterans have joined the broader hardware effort. Tang Tan, a former Apple vice president, became OpenAI’s chief hardware officer after a long career involving iPhone and Apple Watch design.

The trade-secret complaint alleges that OpenAI deliberately solicited confidential information from Apple’s current and former employees. Apple describes the alleged conduct as a shortcut intended to support OpenAI’s entry into consumer devices.

OpenAI has rejected that framing. The company has said it has no interest in other companies’ trade secrets and reportedly maintains that it has seen no evidence supporting Apple’s broader accusations.

A court will not decide the dispute through competing statements. It will examine what information existed, whether Apple protected it, how defendants obtained it, and whether OpenAI used or threatened to use it.

That process can burden a hardware program even before a final ruling. Lawyers may need to inspect design histories, engineering repositories, supplier communications, prototypes, hiring records, and internal messages. Employees can face interviews, depositions, and restrictions on deleting routine records.

OpenAI may also need to demonstrate that disputed features were independently developed. That can require dated design files, decision logs, technical experiments, and documentation showing where ideas originated.

Independent-development records are especially important when teams contain many employees from one competitor. Similarity alone does not prove theft, since experienced engineers often converge on comparable answers to the same constraints. Yet weak documentation can make an otherwise innocent development history harder to establish.

An injunction would present the clearest operational danger. Apple wants the court to stop OpenAI from using Apple information and require the return of confidential materials. The practical impact would depend on how specifically a court defined the protected information.

A narrow order might require deletion or isolation of particular files. A broader order could force OpenAI to examine whether designs, manufacturing methods, supplier choices, or engineering decisions were influenced by disputed material.

The hardware allegations reportedly cover unreleased components, prototypes, confidential designs, and information about unannounced projects. Apple also alleges that OpenAI approached trusted partners while armed with confidential knowledge.

If Apple substantiates those claims, OpenAI could face more than a conventional employment dispute. It might need to rebuild affected workstreams, change suppliers, restrict certain employees, or postpone product milestones while proving clean development.

However, Apple’s requested relief does not guarantee that a device will be blocked. Courts generally expect plaintiffs to connect an injunction to defined trade secrets and identifiable harm. Broad accusations about institutional knowledge are not enough.

The litigation creates timing pressure either way. Consumer devices depend on coordinated hardware, software, manufacturing, and supply-chain schedules. Legal uncertainty can complicate each stage, especially when outside partners worry that their work may become evidence.

OpenAI must now pursue two goals at once. It needs to keep its device program moving while creating a defensible record showing that its engineering belongs to OpenAI, not Apple.

Apple and OpenAI Have Become Partners and Hardware Rivals

The case exposes a relationship that moved from AI cooperation to direct competition without ever becoming a clean separation.

Apple integrated ChatGPT into its software when its own systems could not answer certain user requests. That arrangement gave Apple access to a widely used AI service while allowing OpenAI to reach people through Apple devices.

The partnership made sense when the companies occupied different layers. Apple controlled devices and operating systems. OpenAI supplied a model-based service that users could invoke when Apple’s software needed outside assistance.

OpenAI’s hardware ambitions weaken that division. A dedicated AI device would give the company control over sensors, interface choices, distribution, and the context supplied to its models. Those are areas Apple has spent decades integrating tightly.

The conflict is therefore larger than a disagreement about departing employees. It concerns who will control the main consumer interface for AI.

Smartphones remain the default personal computing device, but AI companies want interactions that require less screen navigation. A device built around voice, vision, memory, or environmental context could challenge the app-based relationship that supports Apple’s platform.

OpenAI does not need to replace the iPhone for that strategy to matter. A companion product could intercept tasks that users currently perform through mobile applications. It could become the first place people ask questions, manage schedules, record ideas, or act on incoming information.

Apple has strong reasons to protect the engineering behind its devices. Its hardware advantage depends on design, component integration, manufacturing processes, supplier relationships, and decisions that remain hidden until products ship.

OpenAI has an equally strong reason to hire people who understand those problems. Model research does not automatically produce a manufacturable consumer device. Hardware requires specialists who can manage batteries, thermal limits, radios, materials, sensors, reliability, and production at scale.

Hiring expertise is legitimate. The dispute begins where expertise allegedly becomes a channel for confidential information.

Apple alleges that Tang Tan encouraged employees interviewing with OpenAI to disclose protected details. Reports on the lawsuit allegations describe claims involving confidential project information and physical components brought into interview settings.

Apple also accuses Chang Liu of retaining an Apple-issued laptop after joining OpenAI. Its complaint reportedly says Liu discovered continued access to cloud storage and downloaded confidential hardware files while already employed by OpenAI.

These remain Apple’s claims. OpenAI and the individual defendants can dispute the descriptions, the secrecy of particular information, and any connection between downloaded material and OpenAI’s products.

The competing narratives are straightforward. Apple presents a coordinated extraction effort directed toward a commercial shortcut. OpenAI presents itself as building with legitimately recruited talent and says it does not want another company’s trade secrets.

The evidence-preservation letters could help determine which account fits the records. Interview instructions, internal reactions, access logs, design timelines, and communications between recruits and managers can reveal whether information moved intentionally.

They can also show whether OpenAI maintained safeguards. Companies hiring from direct competitors often instruct new employees not to bring files, disclose confidential details, or use former employers’ materials. Effective controls usually include documented training, restricted repositories, and escalation procedures.

The existence of a policy would not end the inquiry. Investigators would ask whether managers followed it, whether concerns were reported, and whether OpenAI isolated questionable material once discovered.

Apple’s position carries its own complication. The company must avoid implying that former employees cannot use their accumulated skills elsewhere. California strongly supports employee mobility, and Silicon Valley’s labor market relies on people moving between competitors.

That history includes Apple itself. The company has recruited heavily from rivals to build new capabilities. OpenAI can argue that Apple is trying to turn ordinary knowledge and lawful movement into proprietary territory.

Apple therefore needs a precise case. It must identify information that derives economic value from secrecy and show reasonable measures used to protect it. It must then connect the defendants to acquisition, disclosure, or use through improper means.

The Apple OpenAI legal battle will test more than ownership of individual files. It will test whether a company built around recruiting elite teams crossed the boundary between buying experience and acquiring a competitor’s secrets.

Preservation Letters Are Evidence Requests, Not Proof of Theft

The widening investigation strengthens Apple’s ability to search for evidence, but it does not validate Apple’s accusations by itself.

The number 40 sounds damaging because it implies scale. Yet the figure must be interpreted carefully. Apple reportedly sent notices to about 40 people, while claiming that more than 400 former Apple employees work at OpenAI.

Lawyers frequently cast a wide preservation net. A person can receive a notice because they witnessed an interview, worked on a related project, communicated with a defendant, or handled documents. None of those circumstances necessarily implies misconduct.

The notices can also protect OpenAI employees from later claims that routine deletion destroyed evidence. Once recipients understand the dispute’s scope, they can suspend automatic deletion and preserve relevant devices or accounts.

Apple’s central challenge remains proving misappropriation. A trade secret is not simply information that a company prefers to keep private. It must have economic value because it is not generally known, and its owner must take reasonable steps to protect it.

Apple must identify its secrets without defining them so broadly that they become general engineering knowledge. A product codename, a confidential manufacturing technique, or an unreleased specification could qualify. Broad familiarity with hardware design usually would not.

It must also prove improper acquisition, disclosure, or use. Possession of a file can be significant, but the legal meaning depends on its contents, how it was obtained, why it was retained, and whether anyone used it.

The complaint contains allegations that sound highly specific. Apple says Liu downloaded dozens of confidential hardware-related files, including information about unreleased products and technical specifications. It also says he retained access after leaving through an unusual system flaw.

According to coverage of the network-access claim, Apple alleges Liu discovered the access issue after joining OpenAI and continued retrieving material. Liu and OpenAI will have opportunities to challenge that characterization and the claimed purpose.

Even if an employee improperly retained files, Apple must connect that conduct to OpenAI for its broadest theory to succeed. It will seek evidence showing direction, encouragement, awareness, acceptance, or use by the company.

OpenAI can respond with records showing that it prohibited disclosures, developed designs independently, or never received the materials. It can also contest whether particular information was secret, valuable, or relevant to its work.

The approximately 40 preservation letters could produce evidence for either side. Some records may show coordinated requests for Apple information. Others may show recruiters repeatedly warning candidates not to disclose confidential material.

Another uncertainty concerns the more than 400 former Apple employees. That number establishes a large migration of talent, but it says nothing by itself about when those people left, what roles they held, or whether they work on hardware.

The relevant group may be much smaller. Apple’s selection of approximately 40 recipients could reflect its initial view of who had connections to the challenged conduct. It could also be an early round before discovery identifies additional witnesses.

There is also a risk that aggressive legal tactics chill lawful mobility. Former employees may worry that changing jobs exposes their personal devices, messages, and professional relationships to scrutiny. Current workers may hesitate to interview with a competitor.

Apple has a legitimate interest in protecting qualifying secrets. Yet an overly expansive campaign could appear designed to slow a rival or discourage departures. OpenAI is likely to emphasize that possibility when contesting the scope of discovery or requested relief.

The court will have to balance those interests through concrete disputes. It may decide which repositories must be searched, which employees must testify, and how confidential material should be handled during litigation.

Protective orders can prevent sensitive evidence from becoming public. However, they do not remove the internal burden of collecting, reviewing, and explaining years of technical records.

The case’s public narrative may therefore move faster than its legal findings. Headlines can describe a sweeping espionage operation while the court is still deciding whether Apple identified protectable secrets.

Readers should maintain that distinction. Apple has presented serious and unusually detailed allegations. OpenAI has denied wanting competitors’ secrets. Neither position becomes fact merely through repetition.

The strongest conclusion available now is narrower. Apple believes its evidence justifies a broader investigation, and it is preserving potential records before they disappear. Whether those records prove a coordinated OpenAI strategy remains unresolved.

Three Signals Will Decide What Happens Next

The case will turn on document discovery, the scope of any injunction, and whether OpenAI changes its hardware schedule or development team.

The first signal is what emerges from preserved communications. Apple’s theory depends on connecting individual conduct to a broader OpenAI process. Messages about interview preparation, requested materials, internal codenames, or reactions to confidential files would strengthen that theory.

Records showing clear compliance instructions would weaken it. So would evidence that OpenAI isolated questionable information and documented independent development before Apple filed suit.

The public may not see those records quickly. Trade-secret cases often involve sealed filings and protective orders because the evidence contains the very information a plaintiff seeks to protect. Even so, disputes over discovery can reveal the categories of evidence each side considers important.

The second signal is the court’s treatment of Apple’s requested injunction. Apple wants to stop the defendants from possessing, using, or disclosing its information and to secure the return of confidential materials.

A narrow preservation or return order would protect evidence without necessarily disrupting OpenAI’s program. A broader restriction tied to design work, suppliers, or specific employees would create a larger operational problem.

The wording matters. An order must be specific enough for OpenAI to understand what conduct is prohibited. If Apple cannot define its claimed secrets clearly, the court may resist sweeping relief.

A broad injunction would strengthen Apple’s assertion that the alleged information creates ongoing competitive harm. A denial based on insufficient specificity would weaken the immediate threat to OpenAI, although the underlying case could continue.

The third signal is any visible change to OpenAI’s hardware plans. The company has been preparing consumer devices as a major expansion beyond ChatGPT. Litigation does not automatically stop that work, but schedule shifts can reveal how seriously OpenAI views the risk.

A delayed announcement would not prove that Apple’s allegations are true. Hardware projects slip for many reasons, including component problems, design revisions, manufacturing constraints, and software readiness.

More telling changes would include reassigned leaders, isolated development teams, supplier changes, or a formal review of design provenance. Provenance here means the documented origin and development path of technical work.

The device litigation has already raised questions about whether OpenAI’s launch plans face delays. The answer will depend on whether the dispute reaches core designs or remains focused on identifiable documents and individuals.

Apple’s behavior will offer signals too. Additional defendants or another wave of preservation notices would show that its investigation continues to expand. A detailed amended complaint could reveal new allegations developed from internal records.

OpenAI’s response will be equally important. A motion to dismiss can attack the legal sufficiency of Apple’s claims. A forceful factual response could provide its first detailed account of recruitment safeguards and independent development.

Settlement remains possible, but the companies have conflicting strategic incentives. Apple wants strong protection around its hardware organization. OpenAI wants freedom to employ former Apple workers and launch devices without a cloud over their origins.

Their continuing product partnership adds another layer. ChatGPT’s place in Apple software benefits both parties, but direct hardware competition makes cooperation less stable. Neither company has to end the partnership for mistrust to shape future negotiations.

The dispute also sends a message across the AI sector. Companies hiring concentrated groups from competitors will face increasing pressure to document what new employees bring, what they access, and how teams build products independently.

For engineers, the practical lesson is not that they must leave their expertise behind. It is that files, prototypes, confidential specifications, and internal communications remain the former employer’s protected material. Personal recollection can also create problems when it reveals a specific secret rather than general skill.

For technology buyers and developers, the case matters because it could affect when OpenAI enters consumer hardware and how deeply it controls the AI experience. A delayed device would leave smartphones and existing operating systems in a stronger position.

For knowledge workers, it highlights the growing importance of information provenance. Teams need clear boundaries between personal knowledge, employer records, and material imported into new projects. Those boundaries become harder to reconstruct after a dispute starts.

The Apple OpenAI legal battle has therefore become more than a sensational complaint between two prominent companies. It is an early test of how AI businesses can recruit established hardware teams without inheriting legal exposure alongside their expertise.

The next phase will not be decided by the number of letters alone. Watch whether preserved records connect recruitment to confidential requests, whether a judge restricts OpenAI’s development work, and whether its hardware timetable changes. Those signals will show whether Apple has uncovered a focused employment dispute or a deeper problem inside OpenAI’s device strategy.

What should readers do while those facts develop? Treat Apple’s claims as allegations, not findings, and treat OpenAI’s denial as a position that still requires supporting records. Follow the court’s orders rather than social-media verdicts. Most importantly, watch whether OpenAI can document a clean engineering history while keeping its device program on schedule. That combination, not the rhetoric surrounding approximately 40 legal letters, will determine whether this conflict becomes a temporary distraction or a defining obstacle for OpenAI’s hardware ambitions.