Apple gains expanded access to evidence in OpenAI trade secrets lawsuit
A judge granted Apple broader access to records regarding former employees' AI tool usage and OpenAI's hardware development practices.
The short version
- Apple gained access to records of a former employee's use of ChatGPT, Claude, and other AI tools.
- The court will review details on devices and messaging apps allegedly used to share Apple secrets.
- OpenAI must provide testimony on work involving Apple's battery and power management tech.
- The defendants have 14 days to produce the requested records and respond to questions.
Why it matters
This ruling allows Apple to scrutinize exactly how AI tools were used to potentially move proprietary tech into the hands of competitors. It sets a precedent for how 'trade secret' boundaries are defined when employees move between tech giants and AI startups.
The scope of the lawsuit
Apple is currently engaged in a legal battle against two former employees, Chang Liu and Tang Tan, as well as the companies OpenAI and io Products. The core of the legal complaint involves the misappropriation of trade secrets, which refers to the unauthorized taking of confidential business information. Apple alleges that these individuals improperly accessed or used secret data to help OpenAI advance its own consumer hardware development.
The lawsuit began in July as Apple sought to protect its proprietary technologies from being used by rivals. To achieve this, the company requested a preliminary injunction to stop the defendants from using any alleged secrets. They also asked for expedited discovery, a legal process that allows a party to obtain internal documents and forensic evidence much faster than the standard timeline.
Judge Edward Davila initially granted some of these requests but delayed others because he felt Apple's definition of "Apple Originated Information" was too broad. However, after the parties agreed on a narrower definition, the judge issued a new decision expanding Apple's access to specific evidence. This move allows Apple to dig deeper into how the defendants handled confidential data during their transition to new roles.
Tracking AI tool usage
One of the most significant aspects of the new ruling involves the specific tracking of AI tool usage. The court granted Apple access to records concerning Chang Liu's use of several popular AI platforms, including ChatGPT, Claude, and Codex. This includes data from both personal and corporate devices as well as various accounts used during the period in question.
By reviewing these logs, Apple aims to determine if confidential information was fed into these models to train them or generate new products. The court also ordered the disclosure of details regarding the specific devices, accounts, and messaging apps that Liu and Tan allegedly used. This is intended to map out the communication channels through which secret information might have flowed.
The court is also looking for evidence of any data destruction. Apple requested details on any confidential information that may have been deleted or destroyed, including the specific timing, the method of deletion, and the identity of the person who performed the action. This forensic trail is crucial for establishing whether there was an intentional effort to hide the misappropriation of secrets.
Hardware and recruitment
The expanded discovery also targets the specific technical areas where OpenAI and Apple compete. The court ordered OpenAI and io Products to provide testimony regarding work that involved Apple information related to batteries, power management, and metal finishing. These components are vital to the development of high-end consumer electronics and hardware devices.
Beyond technical specifications, the judge is looking into the human element of the trade secret dispute. The court requested testimony about OpenAI's interviews with Apple employees to see if confidential information was discussed or solicited during the hiring process. This helps the court understand whether the recruitment of talent was a primary vehicle for the transfer of proprietary knowledge.
Furthermore, the court wants to know what steps OpenAI took, or failed to take, to prevent the unauthorized disclosure of Apple's information. This puts the burden on OpenAI to demonstrate that they had adequate safeguards in place. If the company failed to monitor its new hires or secure its internal systems, it could face greater liability in the ongoing litigation.
Limitations on discovery
While Apple won several key requests, Judge Davila did not grant every demand the company made. He denied Apple's requests for specific details about OpenAI's own internal devices and accounts. He also refused to grant access to OpenAI's internal investigations or any records regarding the potential deletion of confidential information by the company itself.
The judge also denied two additional requests for corporate testimony, labeling them as "too broad at this stage." He ruled that these specific questions are better suited for merits discovery, which is the standard, slower process of gathering evidence for the main trial. This means Apple will have to wait longer to pursue those specific lines of inquiry while moving forward with the expedited items.
Despite these limitations, the current order provides Apple with a significant amount of immediate data. The defendants now have a strict 14-day window to produce the requested records and respond to Apple's questions. This accelerated timeline ensures that the most sensitive evidence regarding AI tools and hardware secrets is reviewed quickly.
What this means for the industry
This case highlights the increasing tension between the rapid development of AI and the protection of traditional hardware secrets. As companies like OpenAI move into the hardware space, the line between using public AI tools and using private corporate data becomes blurred. This ruling allows the court to see exactly how those boundaries are being navigated in real-time.
For the tech industry, the focus on AI tool logs like ChatGPT and Claude suggests that these platforms are now central to trade secret litigation. Courts are beginning to treat the inputs and outputs of these models as potential evidence of intellectual property theft. This could lead to stricter internal policies for employees using AI at work.
The outcome of this case will likely influence how tech companies manage the transition of employees between competitors. By forcing OpenAI to testify on their recruitment practices and hardware development, the court is setting a standard for how much oversight a company must provide when hiring from a direct rival. The next 14 days will be critical as the requested records are handed over for review.
Next steps in the case
With the production of records now underway, the focus will shift to the analysis of the data provided by OpenAI and the former employees. Apple's legal team will look for patterns in the AI tool usage and messaging logs to build their case for misappropriation. They will be searching for specific instances where Apple's proprietary battery or power management tech appeared in OpenAI's work.
The defendants must now work to justify their actions and demonstrate that their development efforts were independent of any confidential information. They will need to show that their use of AI tools was for general purposes and did not involve the specific secrets Apple claims were stolen. The evidence regarding the
Sources and links
- Reporting9to5Mac: the article we sourced
Liked it? Say why
Add a line for other readers.
No account needed. We never show your email. Comments with links, ads or abuse are held and not shown. Your comment shows after a quick check.
Listen to the AI Wire headlines
Listen to the AI WireListen to our daily AI shows every morning: on your commute, in the shower, wherever you are
Listen to the daily showsFirst reported by 9to5Mac · Read the original. Written by the SI Agent News Desk from the reporting above.Written with AI from the source above. Every number is checked against the source. Spot a mistake? Corrections.
