In this photo illustration, the OpenAI logo is displayed on a laptop screen on May 20, 2026 in Los Angeles, California.
Justin Sullivan/Getty Images
Three days after Apple filed a federal trade-secret complaint alleging that OpenAI’s hardware chief directed engineers to smuggle unreleased device components to job interviews and that a former systems engineer exploited a network authentication vulnerability to pull confidential files weeks after resigning, OpenAI signaled it has no intention of moving its calendar. According to Bloomberg’s Mark Gurman, writing Sunday, OpenAI still believes it is on track to announce its first consumer device before the end of 2026 and ship it in 2027 — though the company acknowledged that assessment “could still change” as it processes Apple’s claims.
What OpenAI cannot do is make the lawsuit disappear by holding the line on dates. Apple’s complaint, filed July 10 in the U.S. District Court for the Northern District of California, represents something more durable than a publicity move: it is a structured legal instrument that was architected with California’s demanding trade-secret standard in mind, and it is already producing real friction — on recruiting, on supply-chain relationships, and on the pre-IPO disclosure process — without a single court ruling required to do so.
The gap between “timeline unchanged” and “everything is fine” is where this story lives.
What Apple Actually Alleges: Two Defendants, Two Distinct Schemes
Apple’s 41-page complaint, filed under four federal claims and two breach-of-contract counts, names four defendants: OpenAI, its hardware subsidiary io Products, Chief Hardware Officer Tang Yew Tan, and former systems electrical engineer Chang Liu.
Tang Tan spent 24 years at Apple, most recently as VP of product design for the iPhone and Apple Watch. He left in February 2024 to co-found io Products with Jony Ive and others. Apple alleges he ran what the complaint characterizes as a systematic extraction operation: using Apple’s confidential project code names during recruiting interviews, instructing candidates still employed at Apple to bring “actual parts” — batteries, logic boards, system-in-package components — to OpenAI for “show and tell” sessions. He allegedly circulated an internal Apple departure document marked “Need to Know,” which Apple claims he used to coach incoming OpenAI hires on how to avoid the “dreaded walkout” — the revocation of network access that signals an employer has flagged a departing employee for security scrutiny.
Chang Liu’s alleged misconduct is more technically specific. Liu spent eight years at Apple as a senior systems electrical engineer before leaving for OpenAI in January 2026. He did not return his company-issued MacBook when he left, did not respond to Apple’s exit outreach, and did not schedule a confidentiality reminder session. Then, according to the complaint, on or around February 9, 2026 — weeks after his departure and while employed at OpenAI — Liu discovered that an authentication vulnerability in Apple’s internal network still allowed him remote access to shared file servers. His message to a former Apple colleague: “LOL, I found out I can access the [network storage], so funny.” The colleague allegedly replied: “I’m ready.” Apple says Liu proceeded to download dozens of confidential hardware files from Apple’s network, including technical specifications, engineering presentations, and proprietary project data, with one collection exceeding 1,000 pages. He also allegedly sent a message the same day he left Apple: “I still have another computer” — a reference to a second Apple-issued device he planned to use to continue the access.
The complaint further alleges that OpenAI approached a shared contract manufacturer and arranged a demonstration of a proprietary Apple metal-finishing technique — the kind of precision surface treatment that took Apple decades to develop — while misleading the supplier into believing Apple had consented to the demo.
OpenAI’s spokesperson Drew Pusateri issued a statement: “We have no interest in other companies’ trade secrets. We remain focused on building innovative technology that empowers people everywhere.” Apple told MacRumors it is “suing to protect the hard work of its employees.”
Can Apple Actually Block OpenAI’s Device?
This is the question that most coverage has not answered cleanly, and the legal framework requires a clear-eyed answer.
Apple cannot win a product injunction under California law simply by pointing to 400-plus former Apple employees now at OpenAI and arguing their knowledge will “inevitably” flow into OpenAI’s hardware. California courts have explicitly and completely rejected the inevitable disclosure doctrine — the theory that an employee’s background knowledge of a former employer’s trade secrets constitutes threatened misappropriation of those secrets. In Whyte v. Schlage Lock Co., a California Court of Appeal stated the position without ambiguity: “our rejection of the inevitable disclosure doctrine is complete.” The state’s prohibition on non-compete agreements reinforces the same policy: employees are free to move, and their expertise travels with them.
What makes Apple’s complaint potentially viable — and what the draft complaint is carefully structured to demonstrate — is that it does not rely on inevitable disclosure. It relies on alleged specific acts: a retained device, a post-resignation network intrusion, physical components brought to job interviews, a forged-permission approach to a contract manufacturer. Each allegation is a discrete, checkable claim. Under both California trade-secret law and the Defend Trade Secrets Act (18 U.S.C. § 1836), Apple must prove improper acquisition or actual use of the trade secrets — not merely that people who knew Apple’s secrets went to work for a competitor.
This distinction explains why Apple’s complaint reads more like a criminal indictment than a standard civil trade-secret filing: it is building a record of specific conduct because, in California, specific conduct is the only path to injunctive relief.
The bar for a product injunction remains high. Bloomberg Intelligence assessed that Apple is likely to secure targeted preliminary relief tied to OpenAI’s device effort — a narrower prediction about evidence preservation orders and compliance certification requirements, not a full product ban. A court order requiring OpenAI to audit its design decisions for tainted materials, and certify compliance, would still force an internal review that pulls engineering and executive attention away from shipping hardware.
One additional piece of context the current coverage has largely missed: OpenAI is not facing its first preliminary injunction in this space. In a separate case brought by startup iyO — which originally sued OpenAI over the similar name of Jony Ive’s io Products — U.S. District Judge Trina L. Thompson granted iyO a preliminary injunction on April 23, 2026, preventing OpenAI from using the “io” brand while the case proceeds. iyO subsequently added trade-secret misappropriation claims in March 2026. A court has thus already granted a preliminary injunction touching OpenAI’s hardware effort, establishing that the Northern District of California will not reflexively refuse early relief in connected disputes.
If Apple wins a preliminary injunction here — even a limited one — the operational demand on OpenAI’s legal and engineering teams grows significantly.
How the Lawsuit Is Already Causing Damage
Court orders are not required for the complaint to do its work. Three channels of damage are already open.
Recruiting. The complaint’s most immediate effect is a chilling signal to Apple engineers who might have quietly considered OpenAI interviews. Apple’s complaint names specific individuals, documents specific messages, and describes specific procedures Apple used to detect the alleged misconduct. Any Apple employee who considers contacting OpenAI now knows that Apple is surveilling departure patterns closely, that it will name individual defendants in public filings, and that showing up at an OpenAI interview with any material from Apple’s internal systems — or even with notes about what to study — could result in being named in a federal lawsuit. The number of Apple engineers willing to test that boundary has almost certainly dropped.
Inside OpenAI, the complaint forces a different kind of caution: former Apple employees are now likely to grow more careful about what prior knowledge they reference in product discussions, and managers may steer around technical territories that could touch Apple’s IP. The demands of litigation — document preservation, depositions, legal review of engineering decisions — will draw on senior time that had been allocated to shipping hardware.
Supply Chain. The complaint’s allegation that OpenAI approached a shared contract manufacturer with Apple’s proprietary metal-finishing technique — without Apple’s permission — is the supply-chain element most vulnerable to legal remedy. It does not require an employee to have done anything; it requires only that the manufacturer and the relationship itself be examined in discovery. Apple’s leverage over Asian consumer-electronics suppliers is substantial: any manufacturer with significant iPhone assembly contracts will be reluctant to deepen ties with OpenAI if doing so exposes the relationship to litigation.
The supply-chain picture is also more complicated than the initial January 2026 reporting suggested. OpenAI did move manufacturing of its first device — internally codenamed “Gumdrop” and developed by Jony Ive — from Luxshare Precision to Foxconn, with assembly targeted for Vietnam and the United States. That shift, driven by federal security requirements tied to OpenAI’s government infrastructure contracts and by a general preference for a non-China supply chain, insulates the Gumdrop program from some of Apple’s leverage over Luxshare. But subsequent reporting from April 2026 confirmed that OpenAI is running a parallel manufacturing track: Luxshare is the assigned manufacturer for OpenAI’s planned AI phone, a MediaTek Dimensity 9600-based smartphone built on TSMC’s second-generation N2P process and targeted for mass production in the first half of 2027. Luxshare remains in the picture. The Foxconn pivot is real and consequential; it is not a complete exit from manufacturers with high-stakes Apple relationships.
IPO. OpenAI filed a confidential draft S-1 with the Securities and Exchange Commission in May 2026, confirmed publicly June 8. Goldman Sachs, Morgan Stanley, and JPMorgan are leading the offering. Sam Altman initially targeted a September 2026 listing, but by late June, reporting from the New York Times indicated that OpenAI is now leaning toward a 2027 IPO, with Altman calling any valuation below $1 trillion a non-starter. Prediction markets as of July 10–13 were pricing the probability of a 2026 IPO at roughly 18.5 percent, down from 22 percent before Apple’s filing.
The filing creates a direct disclosure obligation. A company heading to public markets must disclose material legal risks, and a complaint that accuses OpenAI’s hardware business of being “rotten to its core by its illegal reliance on misappropriated trade secrets” — in Apple’s words — is precisely the category of language that underwriters must address plainly in an S-1. Apple’s phrasing was not accidental.
Why Apple Chose This Moment
The lawsuit lands at the end of a two-year collapse of what was once the most prominent partnership in consumer AI. In June 2024, Apple integrated ChatGPT into Siri and Apple Intelligence. OpenAI expected the arrangement to generate significant recurring revenue. Relations deteriorated: OpenAI grew dissatisfied with how Apple surfaced ChatGPT within Siri, limiting organic discovery. On January 12, 2026, Apple announced a multi-year partnership with Google to power a rebuilt Siri with a custom Gemini model. ChatGPT was not removed from the Apple ecosystem entirely, but it was demoted from the default position it had occupied.
Reports from before Apple’s filing indicated that OpenAI had itself been weighing legal action against Apple, potentially alleging breach of the 2024 partnership’s integration and promotion commitments. Apple struck first. According to the complaint, Apple sent OpenAI a letter in February 2026 raising its trade-secret concerns. OpenAI did not respond. The lawsuit followed five months later.
For Apple, the strategic calculus is favorable even without a courtroom victory. The talent drain had already reached a point that Apple’s own court filing describes as requiring the rebuild of parts of its iPhone design team. Jony Ive, the designer who shaped the original iPhone, iPod, and Apple Watch, now leads OpenAI’s device work. Apple has responded with retention bonuses for senior engineers. Apple services chief Eddy Cue, in antitrust testimony, warned that AI could mean consumers “may not need an iPhone” in a decade’s time. Apple’s lawsuit is arguably the most efficient instrument available to slow the reckoning it fears: not by winning in court, but by consuming OpenAI’s engineering and executive bandwidth, freezing its recruiting pipeline, and inserting a material legal risk disclosure into the IPO narrative — all without a single favorable ruling required.
What OpenAI’s Device Picture Actually Looks Like
The Bloomberg/Gurman report confirms that OpenAI’s first device remains on track for a 2026 announcement and 2027 commercial release. That product, based on earlier reporting, is the Gumdrop: a compact, screenless audio-first device developed by Jony Ive’s team, manufactured by Foxconn in Vietnam or the United States.
Separate from the Gumdrop, a parallel smartphone program is in development. The AI phone, based on MediaTek’s Dimensity 9600 — a flagship SoC built on TSMC’s N2P process — is targeting mass production in the first half of 2027, with Luxshare as the assigned manufacturer. OpenAI has also been reported to be working on camera-equipped smart speakers.
What Gurman’s July 13 report carefully does not resolve: which specific product OpenAI will announce this year. Earlier reporting identified the smart speaker as the first device likely to actually ship, with the Gumdrop and the AI phone on longer development timelines.
The lawsuit’s product-level risk, in the near term, is not a launch cancellation — California courts will not grant that without years of evidentiary proceedings. The risk is a court-ordered audit of which design decisions may have been informed by the materials Liu allegedly downloaded or the intelligence Tan allegedly gathered in “show and tell” sessions. That audit, if ordered, touches every product on the roadmap.
How Apple Wrote Its Complaint to Win in California
One piece of legal history that has not appeared in current coverage is worth noting. In 2020, Apple was the defendant in a trade-secret case: Hooked Media Group, Inc. v. Apple Inc. A startup accused Apple of misappropriating its recommendation algorithm after Apple hired away three of Hooked’s engineers, including its chief technology officer. The California Court of Appeal ruled against Hooked, holding that Apple did nothing wrong — because knowledge of trade secrets, transferred through employees who move to a new employer, does not constitute misappropriation under California law. The court was explicit: a plaintiff must prove improper acquisition or actual use, not just that the employee carried knowledge across employers.
Apple is now the plaintiff making allegations structurally similar to what Hooked accused it of — but with one critical difference that the Hooked precedent precisely identifies: Apple’s complaint is built not on what OpenAI’s recruits knew, but on what they allegedly physically took, accessed without authorization, and physically delivered to OpenAI. That is not an inevitable-disclosure case. It is an actual-conduct case. The Hooked ruling defines the bar Apple’s complaint must clear; Apple’s legal team, having won under that bar as a defendant, knows exactly where it sits.
OpenAI’s defense will likely attack the chain of causation — that even if Liu downloaded those files, Apple cannot prove the specific documents informed specific design decisions in OpenAI’s hardware. That is a long, expensive, discovery-dependent argument. It is also the argument most likely to extend the litigation’s uncertainty over OpenAI’s product plans rather than resolve it quickly.
What to Watch For
Several near-term developments will clarify the lawsuit’s actual trajectory.
A motion for a preliminary injunction is likely. Apple indicated in its filing that it intends to move promptly for preliminary relief. The question is whether that relief would be narrowly targeted — evidence preservation, use-cessation certification — or broad enough to touch product design decisions directly. Bloomberg Intelligence assessed that targeted preliminary relief is the more likely outcome.
The 2026 device announcement itself becomes a test. If OpenAI holds to its stated plan, the announcement will reveal whether the product’s form factor, feature set, or manufacturing approach shows signs of adjustment — either to distance itself from Apple’s specific allegations or to demonstrate independence from the design elements at issue.
The Foxconn-Luxshare split requires continued scrutiny. Conflicting supply-chain accounts — some sources maintaining Luxshare’s involvement in the AI phone program while Foxconn handles the Gumdrop — suggest that OpenAI’s manufacturing picture is more complex than the initial January 2026 reporting captured. The Apple complaint’s allegation about the shared contract manufacturer may identify the specific manufacturing relationship most legally exposed.
The IPO timeline carries a disclosure obligation that does not disappear regardless of when OpenAI lists. The Apple complaint’s language will appear in the S-1’s risk factors. How OpenAI characterizes the litigation’s potential impact on its hardware strategy — a business that one complaint has now described in terms no underwriter can ignore — will be one of the most closely watched sections of any prospectus.
Frequently Asked QuestionsCan Apple stop OpenAI from launching its AI device?
In the near term, an outright launch ban is not a realistic legal outcome. California courts have explicitly rejected the “inevitable disclosure” doctrine that would let Apple seek an injunction simply because its former engineers now work for OpenAI. To win product-level injunctive relief, Apple would need to trace specific stolen materials — the files Liu allegedly downloaded, the intelligence gathered in “show and tell” sessions — to specific design decisions in OpenAI’s hardware. That kind of causation proof takes years to build in discovery. What Apple can realistically seek, and what Bloomberg Intelligence says is likely, is targeted preliminary relief requiring OpenAI to preserve evidence, certify it has stopped using the alleged trade secrets, and isolate disputed materials during the case — a narrower order that still forces an internal audit with real operational consequences.
What exactly is the network authentication exploit Apple describes?
According to Apple’s complaint, Chang Liu discovered that after he resigned from Apple in January 2026, an authentication vulnerability in Apple’s network infrastructure still allowed his credentials to access internal shared file servers. Rather than reporting the vulnerability to Apple, he allegedly exploited it: accessing and downloading dozens of confidential hardware files — including engineering presentations, technical specifications, and proprietary project data — weeks after his employment had ended. Apple also alleges Liu did this from a former colleague’s Apple-issued device, making her an unwitting conduit. The complaint describes one collection of downloaded files as exceeding 1,000 pages. This is a specific category of IT security failure — credential persistence after offboarding — distinct from a hack by an outside party.
How does Apple’s lawsuit threaten OpenAI’s IPO?
OpenAI filed a confidential draft S-1 with the SEC in May 2026 and, as of late June, appears to be targeting a 2027 listing at a $1 trillion valuation floor. The Apple lawsuit creates a direct material-risk disclosure obligation in the prospectus. Apple’s complaint language — characterizing OpenAI’s hardware division as “rotten to its core by its illegal reliance on misappropriated trade secrets,” in Apple’s filing — is precisely the kind of allegation that underwriters must address plainly in risk factors. Prediction markets moved noticeably after the filing: the probability of a 2026 IPO fell from roughly 22 percent to 18.5 percent within days. Beyond the disclosure requirement, the lawsuit adds to an already crowded legal horizon: 42 state attorneys general launched a coordinated probe in June 2026, and the Musk appeal to the Ninth Circuit remains pending.
Why is OpenAI still using Luxshare if it moved to Foxconn to avoid Chinese suppliers?
OpenAI’s manufacturing pivot in January 2026 moved its first device — the “Gumdrop” designed by Jony Ive — from Luxshare to Foxconn, with assembly in Vietnam or the United States. That shift was driven by federal security requirements tied to OpenAI’s government infrastructure contracts and by a preference for non-China production. However, subsequent April 2026 reporting confirmed that Luxshare has been assigned to a separate, parallel product: OpenAI’s planned AI phone, based on a MediaTek chip built on TSMC’s N2P process. OpenAI appears to be running a two-manufacturer model — similar to how Apple itself eventually split iPhone production between Foxconn and Luxshare to reduce single-supplier dependency. The strategic logic is sound; the complication is that Luxshare’s dual role as a current OpenAI supplier and a longtime Apple supply-chain partner is precisely the kind of relationship Apple’s complaint targets.