Apple’s legal battle with OpenAI has entered a more serious phase after forensic evidence from a former Apple engineer’s MacBook allegedly revealed the use of confidential Apple engineering information at OpenAI.
In a supplemental court filing on August 31, Apple described the findings as “shocking evidence.” The company says former Apple senior system electrical engineer Chang Liu downloaded a confidential circuit schematic after joining OpenAI, used it for engineering simulations, and trained an AI agent to work with the simulation software.
Apple also alleges that Liu later instructed an OpenAI colleague to destroy evidence after learning that Apple was investigating his activities.
The claims have not been proven in court. OpenAI has denied wrongdoing and has asked the court to dismiss Apple’s lawsuit.
The latest development could nevertheless become a critical turning point in the Apple OpenAI trade secrets lawsuit, because Apple is using the forensic findings to argue that it needs accelerated access to additional devices, documents and witnesses.
What Apple says it found on Liu’s MacBook
According to Apple’s supplemental filing, Liu’s lawyers provided his MacBook for forensic examination. OpenAI provided the device to Apple on August 21 as part of the litigation process.
Apple says its initial analysis produced four significant findings.
First, Liu allegedly downloaded a confidential Apple power-converter circuit schematic from Apple’s third-party cloud storage on March 7, 2026.
That date is important. Liu had already left Apple for OpenAI in January.
Apple alleges that the file contained not only the circuit schematic but also simulation input data used to model the circuit’s operating modes. Reuters reported that Apple claims Liu subsequently used the proprietary information while working at OpenAI.
Second, Apple says Liu actually ran the confidential schematic through LTspice, an electrical-engineering simulation application.
Third, Apple alleges that Liu and others at OpenAI knew that he still had access to Apple’s third-party cloud storage.
Fourth, Apple says that after Liu learned about Apple’s internal investigation, he instructed an OpenAI colleague, identified in reports as Yu-Ting Peng, to destroy evidence. Apple says the colleague confirmed that she would comply.
Those allegations make the dispute considerably more serious than a conventional employee-access or data-retention dispute.
The AI-agent element may be the most consequential
The most technologically significant part of Apple’s allegations concerns what Liu reportedly did with the confidential circuit information.
Apple says Liu used the schematic in LTspice while working on power-conversion engineering. Messages cited in reports about the filing indicate that Liu also described teaching an AI agent to operate LTspice, inspect its results and adjust simulation parameters.
In one exchange, Liu reportedly said that his AI agent had learned to run LTspice, examine results and tune a compensation parameter.
The significance extends beyond the alleged use of a single confidential file.
Apple argues that feeding proprietary engineering information into an AI agent creates a different kind of trade-secret risk. If an AI system learns from confidential information and incorporates that knowledge into its subsequent operations, determining exactly where the information went and whether it can be recovered may become considerably harder.
That is particularly important in an era of agentic AI.
Traditional software generally performs a defined operation on a defined dataset. AI agents can instead interact with tools, examine outputs, modify parameters and repeat a task. The distinction matters when the underlying information is alleged to be a protected trade secret.
Apple is therefore presenting the alleged AI-agent activity as evidence of active use rather than merely unauthorized possession.
Why the MacBook evidence matters
The device itself appears to have become an important piece of Apple’s evidentiary strategy.
Apple says Liu used the confidential schematic on a Mac mini and that the Mac mini later synchronized information to the MacBook that was eventually provided for forensic examination.
That alleged chain could explain how investigators identified activity that occurred on another device.
Apple is now seeking access to the Mac mini and other devices that remain in Liu’s possession. The company argues that the MacBook evidence demonstrates why broader and faster discovery is necessary.
This is also why the Apple OpenAI trade secrets lawsuit is increasingly becoming a battle over digital forensics.
The dispute is no longer limited to what employees allegedly knew or what information may have been discussed during recruitment. It is moving toward questions about device histories, cloud access, synchronization records, engineering files, AI-agent activity and the preservation of electronic evidence.
Apple says evidence may have been destroyed
The evidence-destruction allegation could prove particularly important in the litigation.
Apple says Liu learned about its investigation and subsequently sent instructions to an OpenAI colleague to destroy evidence.
If ultimately established, such conduct could create additional legal consequences beyond the underlying trade-secret allegations.
For now, however, this remains an allegation contained in Apple’s filing.
Apple is using the claim to support its request for expedited discovery. Its argument is straightforward: if evidence is being destroyed or potentially altered, waiting for ordinary discovery could increase the risk that relevant information disappears.
The company has therefore asked the court to accelerate the process of gathering evidence.
The dispute over evidence preservation is especially significant because digital information can exist across multiple devices and services. A file deleted from one computer may remain in cloud storage, backups, synchronized devices, messaging systems or forensic artifacts.
That makes the forensic examination of Liu’s devices potentially more consequential than the contents of any single document.
More than 400 former Apple employees are now at OpenAI
The latest allegations sit within a much broader dispute over talent.
Apple’s original lawsuit says more than 400 former Apple employees now work at OpenAI. Apple has used that figure to argue that OpenAI has extensive access to people who possess knowledge of Apple’s proprietary technologies and development processes.
The number itself does not establish wrongdoing.
Employees routinely move between technology companies. Engineers also carry legitimate professional knowledge and experience from previous employers.
The legal question is where the boundary lies between general expertise and protected confidential information.
That distinction becomes especially difficult in highly specialized engineering fields.
An engineer can legitimately use knowledge gained through years of experience. But using a former employer’s confidential circuit design, simulation inputs, source files, unreleased product specifications or protected manufacturing information can raise a very different set of legal questions.
Apple’s case seeks to establish that some of the information involved crossed that line.
OpenAI disputes Apple’s case
OpenAI has consistently rejected Apple’s allegations.
In an August response, OpenAI said Apple was getting the situation wrong and argued that Liu’s access to Apple information after leaving the company was connected to Apple’s own employees asking him for assistance. OpenAI also argued that residual access to Apple’s systems reflected weaknesses in Apple’s management of access after employees departed.
OpenAI has also argued in court that Apple’s complaint does not identify its alleged trade secrets with sufficient specificity and does not adequately establish misappropriation, harm or other elements required for the claims.
The company has asked the court to dismiss the lawsuit.
As of August 31, OpenAI had not publicly issued a separate response specifically addressing Apple’s newest forensic allegations, according to Reuters.
That distinction matters.
The new MacBook findings have been presented by Apple. They have not yet been independently adjudicated by the court.
A bigger battle over AI hardware
The case is unfolding against an increasingly competitive AI-hardware landscape.
Apple’s July lawsuit accused OpenAI, Liu and former Apple executive Tang Yew Tan of benefiting from Apple’s confidential information as OpenAI develops its own consumer hardware ambitions.
OpenAI acquired io Products, the hardware company associated with former Apple design chief Jony Ive, in 2025. The lawsuit therefore involves not only employee mobility but also competition over future consumer devices.
That makes Apple’s allegations strategically significant.
Hardware development depends on years of accumulated engineering knowledge. Power management, component design, thermal engineering, manufacturing processes and supplier relationships can represent substantial competitive advantages.
The alleged use of a confidential power-converter schematic therefore has significance beyond one engineering experiment if Apple’s account is eventually substantiated.
What happens next in the case?
The immediate battle is over discovery.
Apple wants accelerated discovery and broader access to devices and information. It argues that the forensic findings justify moving faster than the normal litigation timetable.
OpenAI, meanwhile, is seeking dismissal and continues to characterize Apple’s claims as unsupported.
The court will ultimately have to determine which allegations can proceed and what evidence can be examined.
A hearing on Apple’s requests is scheduled for October 1, according to MacRumors.
The next phase could therefore focus heavily on forensic evidence.
Investigators may need to establish when files were accessed, what devices were involved, whether confidential material was actually used in OpenAI work, what an AI agent processed, whether any information was retained, and whether evidence was deliberately destroyed.
Those questions could determine whether Apple’s latest filing represents a major breakthrough or simply another contested chapter in a highly aggressive legal dispute.
Why this case matters for enterprise AI
The Apple OpenAI trade secrets lawsuit raises a much broader question for every company deploying AI agents: what happens when confidential corporate information becomes part of an AI-assisted workflow?
Traditional data-loss prevention systems were designed around files, databases, email and removable media.
AI agents complicate that model.
An agent can potentially read a document, invoke a software tool, modify parameters, inspect results and produce a new output. The original confidential file may never appear in the final output, yet its information could influence the process.
That creates new governance requirements.
Companies will increasingly need clear controls around which AI agents can access engineering repositories, design systems, customer information, source code and other sensitive intellectual property.
They will also need auditable logs showing what information an agent accessed and what actions it performed.
The Apple dispute illustrates why those controls are becoming a board-level issue rather than simply an IT-security concern.
The allegations are serious, but the verdict is still pending
Apple’s latest filing unquestionably raises the stakes.
Its allegations connect a former engineer, confidential engineering data, an electrical simulation tool, an AI agent, cloud access and alleged evidence destruction. The company says the evidence came from forensic analysis of a device supplied during the litigation.
But allegations are not findings of fact.
The court has not determined that Liu misappropriated Apple’s trade secrets, that OpenAI used Apple’s confidential information, or that evidence was deliberately destroyed.
OpenAI continues to dispute Apple’s characterization of events and is seeking dismissal.
The significance of the Apple OpenAI trade secrets lawsuit will therefore depend on what happens when the evidence is subjected to discovery, expert analysis and judicial scrutiny.
For the technology industry, however, the case is already a warning.
The next generation of corporate trade-secret disputes may not revolve around a stolen document sitting on a hard drive.
They may revolve around what an AI agent saw, what it learned, what tools it operated, what outputs it influenced—and whether those actions can be reconstructed months later from forensic evidence.