OpenAI is pushing back against Apple’s trade-secret allegations with a pointed argument: Apple employees allegedly continued asking former engineer Chang Liu for technical help after he had left the company.
The company says it published messages showing Apple staff contacting Liu about technical decisions, schematics, internal folders and colleagues who could help locate information. OpenAI presents those exchanges as evidence that Apple’s own offboarding and access controls were less disciplined than its legal claims suggest.
That is a potentially damaging detail for Apple, but it is not a complete answer to the case. The messages have not been independently authenticated, and even if they are genuine, they address only a narrow part of the dispute. They may raise questions about how Apple handled Liu’s departure without disproving allegations that confidential information was improperly accessed or that OpenAI encouraged incoming employees to bring proprietary material with them.
What OpenAI says the employee chats show
OpenAI says Liu’s final day at Apple was January 22, 2026. In its account, Apple employees continued contacting him almost immediately after his departure rather than treating the employment relationship as cleanly severed.
One message dated January 27 appears to show an Apple employee asking Liu for a technical assessment and describing him as the best person to consult despite his no longer working there. Liu reportedly responded with information about decisions made inside Apple. OpenAI has offered the exchange as part of its defense, but the message and the surrounding context have not been independently verified.
OpenAI says the same employee contacted Liu again on February 14 with a question involving schematics. A further exchange dated March 5 appears to place Liu in a group conversation with several Apple employees. In that thread, he allegedly directed participants toward internal folders and other contacts before asking to be removed because the situation was “highly irregular.”
If accurately presented, the exchanges would suggest Apple employees themselves blurred the boundary between a departed colleague and an active member of the organization. They could also point to weaknesses in employee offboarding security, including unclear communication rules and possible continued access to company systems. OpenAI characterizes this kind of post-employment access as a recurring problem at Apple, but that broader characterization remains OpenAI’s assertion rather than an established finding.
The distinction matters. An employee asking a former colleague where a file is stored is not automatically evidence that the former employee was authorized to access or disclose it. At the same time, a company that continues treating a departed engineer as an informal technical resource may face harder questions about how consistently it protected the information it later described as confidential.
The chats challenge Apple’s procedures, not its entire case
OpenAI’s rebuttal is strongest when it focuses on Apple’s internal conduct. The alleged exchanges could complicate any simple narrative in which Liu left Apple, severed all contact and then independently accessed information that Apple had carefully placed beyond his reach.
But the messages do not establish what information Liu possessed when he joined OpenAI, how he obtained it or whether anyone at OpenAI asked him to retain it. They also do not resolve Apple’s broader allegation that departing employees were encouraged to collect files or specifications during their notice periods.
| Issue | What OpenAI’s response addresses | What remains unresolved |
|---|---|---|
| Contact after Liu’s departure | Apple employees allegedly continued asking Liu for technical assistance. | Whether every message is authentic, complete and presented with its full context. |
| Access management | The chats may indicate weak boundaries around former employees and internal information. | What systems or files Liu could access and whether any access was authorized. |
| Trade-secret allegations | Apple’s own practices could affect how its handling of confidential information is evaluated. | Whether protected information was taken, shared or used improperly. |
| OpenAI’s conduct | OpenAI says it did not welcome proprietary information from competitors. | Whether managers encouraged, knew about or benefited from any improper transfer. |
This is why the chat release works better as a credibility challenge than as a definitive defense. It shifts attention toward Apple’s security processes and the conduct of its employees, but it does not directly rebut every allegation involving recruitment, file retention or the use of proprietary information.
The dispute reaches beyond one former engineer
Apple’s complaint, as characterized in the dispute, alleges systematic trade-secret theft and claims Liu improperly accessed confidential information after joining OpenAI. Those allegations have not been independently established and will depend on evidence beyond the selected employee messages released by OpenAI.
The complaint is also said to place Liu within a much larger movement of talent between the companies, claiming that more than 400 former Apple employees work at OpenAI. That figure has not been independently verified. OpenAI’s expanding hardware ambitions provide important context for the conflict, particularly its work connected to io Products and former Apple design leader Jony Ive.
The overlap between Apple’s former workforce and OpenAI’s hardware operation helps explain why the dispute carries weight beyond a routine employee departure. Apple has spent decades developing closely integrated consumer hardware, while OpenAI is pursuing devices that could create a new way for people to interact with artificial intelligence. The legal fight therefore touches not only individual conduct but also the movement of specialized expertise between direct or potential competitors.
Another important figure is Tang Tan, the longtime Apple executive who later became involved in OpenAI’s hardware efforts. OpenAI rejects the allegations concerning Tan and says he repeatedly instructed his team not to bring or use proprietary information belonging to other companies. That account represents OpenAI’s position; it does not independently establish how every employee behaved or what information may have moved between organizations.
OpenAI is also attacking Apple’s handling of the dispute
Beyond the employee chats, OpenAI alleges that Apple’s outside counsel made significant mistakes during early communications between the companies. OpenAI says a lawyer contacted the wrong person after confusing two surnames and later described a phone conversation with OpenAI’s general counsel that OpenAI says did not occur.
Apple reportedly acknowledged the errors after OpenAI raised them. OpenAI further maintains that the detailed allegations eventually included in the complaint were not presented during those earlier communications. In its telling, Apple’s lawyer initially indicated that the company was trying to sort out potential issues, followed by roughly five months without further contact before the case was filed.
These details have not been independently verified, but they support OpenAI’s effort to portray Apple’s approach as rushed and unnecessarily aggressive. If accurate, the errors could weaken Apple’s broader presentation by showing that important early communications were handled imprecisely. They would not, however, determine whether trade secrets were actually acquired or used.
The disagreement over the correspondence also illustrates the limits of a public rebuttal assembled by one side. OpenAI has selected material that supports its account, just as Apple’s complaint presents Apple’s theory of the case. Readers should separate evidence that exposes an inconsistency from evidence that resolves the underlying allegation.
Verdict: A sharp rebuttal with a major gap
OpenAI’s response creates a legitimate problem for Apple’s narrative. If Apple employees continued requesting technical help from Liu, directed him into internal discussions or left him with access after his departure, Apple may have to explain why its own boundaries were so porous. Those questions are especially relevant when Apple is arguing that the information at issue was tightly controlled.
The chat logs also give OpenAI an effective public argument: Apple cannot easily portray every post-employment interaction as suspicious if its own staff kept initiating those interactions. The alleged mistakes by outside counsel add another line of attack against Apple’s preparation and credibility.
Still, OpenAI’s evidence does not clear the company of the central allegations. Weak offboarding at Apple and improper conduct at OpenAI could both exist. Showing that Apple employees contacted Liu does not establish that every response was permissible, nor does it answer whether unrelated files or specifications were retained, transferred or used.
The most defensible reading is that OpenAI has exposed questions Apple may need to answer, not that it has disproved the case. Until the messages are authenticated, placed in full context and weighed alongside evidence about access, recruitment and the handling of confidential material, they remain a forceful but incomplete defense.