01 / THE FILINGWhat Apple actually alleges
On Friday, July 10, 2026, Apple filed a trade secret complaint in the U.S. District Court for the Northern District of California against OpenAI, its hardware subsidiary io Products, and two former Apple employees: Tang Tan, a longtime Apple product-design vice president who is now OpenAI's chief hardware officer, and Chang Liu, a senior systems electrical engineer who spent eight years at Apple before joining OpenAI in 2026. The filing's own framing is sweeping — misconduct “at every level, from members of its Technical Staff to its Chief Hardware Officer” — and it lands amid a hardware race, as OpenAI prepares its first consumer device. Apple says it raised its concerns with OpenAI in a letter in February 2026 and received no response.
One thing to hold onto for the rest of this article: these are allegations. None has been tested in court, the defendants will answer them, and the case will take years. We're not here to referee it. We're here because the specific scenarios the complaint describes — an unreturned laptop, live network access after departure, hardware components walking out, an offboarding checklist turned against its owner — are the exact scenarios IT asset management, information security, and IT asset disposition programs exist to prevent. Apple's lawyers have, in effect, published a control-failure taxonomy. It would be a waste not to learn from it.
02 / THE LAPTOPThe device that never came back
The allegation. Per the complaint, Liu failed to return his Apple-issued laptop after leaving for OpenAI — and then used that laptop to access Apple's network and download dozens of confidential files, including technical specifications, engineering presentations, and proprietary data on unannounced products.
The control it implicates. Two, actually, and they fail together. The first is verified device recovery at offboarding. Most organizations have a return-your-equipment step; far fewer verify it — per serial number, reconciled against the asset register, with a named owner and an escalation path when something doesn't come back. An unreturned device that nobody chases isn't a missing line item. It's an endpoint you no longer control, holding whatever it held on the employee's last day, for as long as it takes anyone to notice.
The second is access revocation timing. The downloads the complaint describes weren't hacking — they were allegedly performed with credentials and connectivity that remained valid after the employment relationship ended. Access should die on the separation date, not on the equipment-return date, and never on the “whenever IT gets to the ticket” date. If your deprovisioning depends on the same checklist that recovers the hardware, one missed step takes both controls down at once — which is precisely the combination the filing describes.
There's an ITAD corollary here. The same reconciliation machinery that catches an unreturned laptop is what makes end-of-life defensible: serialized intake, chain of custody, and a per-device record that the asset was recovered, sanitized, or destroyed. A fleet you can't reconcile at offboarding is a fleet you can't reconcile at disposition either — the gap just surfaces later, usually at the worst possible time. (And if recovered devices come back BIOS-locked or MDM-enrolled, that's its own preparation problem.)
03 / THE PARTSHardware “show and tell”
The allegation. The complaint alleges Tan directed job candidates who still worked at Apple to bring “actual parts” — batteries, logic boards, system-in-package components — to their OpenAI interviews for “show and tell” sessions.
The control it implicates. Component-level physical asset control. Most asset registers track devices; almost none track the components inside them, and prototype parts in an R&D environment are exactly the class of physical asset that carries trade-secret value with no serial number attached. If a logic board can leave the building in a jacket pocket and nothing reconciles, the control gap isn't hypothetical — the complaint describes it being exploited as a recruiting ritual.
For most organizations the lesson scales down sensibly: you don't need to serialize every screw. You need to know which classes of physical asset would hurt if they left — prototype hardware, drives pulled from decommissioned systems, engineering samples — and put custody controls on those. It's the same discipline ITAD applies at end-of-life: a pulled hard drive awaiting destruction is a component, not a device, and it's precisely when media leaves its chassis that the documented disposal breaches tend to happen. Custody has to follow the part, not just the parent asset.
04 / THE FILESSelf-emailing, and the access that allowed it
The allegation. Per the filing, Tan began emailing himself information about Apple's suppliers before resigning. Separately, Liu allegedly shared Apple confidential information with other Apple employees who were applying to OpenAI — advising at least one on what to study before an interview.
The control it implicates. Restricted file access, and monitoring where restriction isn't possible. Both allegations describe insiders using access they legitimately held. That's what makes insider risk categorically different from external threat: you can't firewall an employee out of their own job. What you can do is three things. First, least privilege — access scoped to role, so a departing employee's reachable blast radius is bounded by what their work actually required. Second, heightened monitoring during notice periods — outbound email volume, personal-account forwarding, bulk downloads, and removable-media use are all measurable, and the interval between resignation and departure is when they matter most. Third, need-to-know as culture, not just configuration — the coaching allegation describes confidential information moving person-to-person, which no DLP rule catches. Classification only works when people know what's classified and why it stays inside.
None of this requires surveillance theater. It requires accepting that the notice period is a control window, not a formality — the same way the offboarding checklist is.
05 / THE PLAYBOOKThe offboarding doc that switched sides
The allegation. The complaint alleges Tan retained or obtained an internal Apple “Need to Know” offboarding document and circulated it at OpenAI — using it to teach incoming hires how to get around Apple's exit security checks.
The control it implicates. Process integrity: your security procedures are themselves confidential assets. An offboarding checklist is a map of your controls — what gets checked, in what order, by whom, and by omission, what doesn't get checked. In the wrong hands it inverts from control to countermeasure. The fixes are unglamorous but real: classify and access-restrict security procedure documents like the sensitive assets they are; assume any long-tenured departure knows the process and vary what you verify; and periodically red-team your own offboarding — walk a hypothetical bad-faith departure through it and see what a person holding your checklist could carry past it.
There's a quieter lesson too: a process document is only as good as the verification behind it. A checklist that gets checked off is not the same as assets that got recovered and access that got revoked. If your offboarding produces no evidence — no reconciliation record, no revocation timestamp — then whether it was followed is a matter of memory, and memory is what ends up in depositions.
06 / THE SCALEFour hundred departures
The allegation. The complaint states that more than 400 former Apple employees now work at OpenAI, and frames the alleged misconduct as institutional — coordinated “at every level” rather than a rogue individual.
The control it implicates. Offboarding as a system, not an event. At 400 departures to a single competitor, offboarding isn't an HR ceremony — it's a control surface under sustained, adversarial load. The statistical reality every organization should internalize: if enough people leave, some departures will test your process, and the process has to hold without depending on goodwill, on the departing employee's honesty, or on any single person remembering to do their step. That means the controls above — verified recovery, automated day-one revocation, notice-period monitoring, procedure-document classification — have to run on rails. A process that works because everyone involved is acting in good faith is not a control. It's a hope with a checklist attached.
This is also where the talent-market framing matters for everyone else: your employees move to competitors too, in every industry, at every scale. The complaint just describes the phenomenon with unusually expensive stakes.
07 / THE LESSONSSeven allegations, one control map
| What the complaint alleges | The control it implicates |
|---|---|
| Company laptop never returned after departure | Verified per-serial device recovery at offboarding, reconciled to the asset register, with escalation |
| Confidential files downloaded via that laptop after departure | Access revocation effective on the separation date — independent of equipment return |
| Hardware components brought to interviews (“show and tell”) | Component-level physical custody for sensitive asset classes, not just parent devices |
| Offboarding document repurposed to evade exit checks | Security procedures classified and access-restricted as confidential assets; offboarding red-teamed |
| Supplier information self-emailed before resignation | Least-privilege access plus heightened exfiltration monitoring during notice periods |
| Confidential information shared to coach other candidates | Need-to-know classification enforced as culture, not just configuration |
| 400+ departures to one competitor; alleged institutional pattern | Offboarding as an automated, evidence-producing system that doesn't depend on goodwill |
Reduced to practice, a defensible program looks like this:
- Every issued device is serialized in an asset register, and offboarding reconciles against it — a device is “recovered” when the serial is verified back in custody, not when the checkbox is ticked.
- Access revocation is automated to trigger on the separation date, decoupled from hardware return.
- Notice periods carry heightened monitoring of bulk downloads, personal-account forwarding, and removable media.
- Sensitive component classes — prototypes, pulled media, engineering samples — carry custody controls of their own.
- Security procedure documents are classified, access-restricted, and periodically revised on the assumption that departed employees know the old version.
- End-of-life is closed with the same rigor it opened: certified sanitization or destruction, per-device certificates naming the NIST 800-88 method, unbroken chain of custody.
Two working companions to this list live on the hub: the Offboarding Gap Diagnostic scores your current posture against these controls in two minutes, and the Vault’s Employee Offboarding IT Asset & Access Checklist is the per-departure working document that operationalizes them.
None of it is exotic. That's the recurring theme of every incident we've written up, from the disposal breach case files onward: the failures are ordinary, and so are the controls that would have prevented them. What the Apple filing adds is a reminder of where in the lifecycle the ordinary failures cluster — at the exits.
Close the loop on the assets you already control
The endgame of every control above is an asset you can account for — recovered, reconciled, and retired with evidence. CyberCrunch delivers the last mile: serialized chain of custody, R2v3 / NAID AAA audited processing, and per-device certificates of destruction. If you handle CUI, our CMMC disposition checklist maps the requirements.
Apple v. OpenAI FAQ
Where do these facts come from — and are they proven?
From Apple's complaint, filed July 10, 2026 in the U.S. District Court for the Northern District of California, via the filing's own language and contemporaneous coverage. They are allegations: the defendants haven't litigated them, and this article doesn't assess the merits. The asset-management lessons stand either way, because the scenarios described are ones every IT organization already has to defend against. (Informational only, not legal analysis.)
What does a trade secret lawsuit have to do with ITAD?
ITAD is the tail end of asset management, and the complaint's most concrete allegations are asset-lifecycle failures: a laptop allegedly never recovered, components allegedly leaving the building. The machinery that makes disposition defensible — serialized inventory, verified recovery, chain of custody, documented end-of-life — is the same machinery that closes these gaps. A device you can't account for is the same problem whether it's headed to a competitor or a landfill.
Would encryption or MDM have prevented the alleged laptop incident?
They help, but neither substitutes for recovery and revocation. Encryption protects data at rest against an outsider — not against the authorized user still holding valid credentials, which is what the complaint describes. MDM enables remote lock and wipe, but only if someone triggers it, which requires the process to notice the device is missing. The controls that map directly to the allegation are verified per-serial recovery and access revocation effective on the departure date.
What should we actually implement?
Six controls cover the described failure modes: verified per-serial device recovery with escalation; access revocation automated to the separation date; heightened exfiltration monitoring during notice periods; least-privilege access scoped to role; component-level custody for sensitive hardware classes; and classification of your own security procedures as confidential assets. For retired equipment, certified destruction with per-device certificates closes the loop — the checklist above operationalizes all of it.