Apple has filed for a preliminary injunction against OpenAI to block development of products based on Apple technologies. The original complaint named two former employees: senior systems engineer Chang Liu and Chief Hardware Officer Tang Yew Tan. The latest motion references 11 additional former employees potentially involved: one took screenshots of confidential documents before an OpenAI interview, another discussed unannounced products in a meeting with two departing colleagues. Since the complaint, several ex-Apple employees now at OpenAI have contacted Apple to return work devices never handed back when they left.

OpenAI responds that the injunction is “based on false information and entirely unnecessary,” and denies having or wanting Apple trade secrets.

Why this matters to you. If you’re deploying AI in your organization, governance covers what happens next: what becomes of your data when someone with access moves on. The same controls protecting an agent in production (logging, least privilege, access tracking) apply when a colleague with access to your AI systems joins a competitor. The Apple-OpenAI case is the highest-profile version of a problem every company already faces, and the Univé case study on ChatGPT Enterprise put it first: rules come before use cases.

The details

Apple’s case against OpenAI rests on a concrete hypothesis: employees with access to unannounced hardware and software projects brought that know-how to a leading competitor. The two named in the original complaint, Chang Liu and Tang Yew Tan, now work at OpenAI and io, the device design startup co-founded by former Apple designer Jony Ive.

August’s motion widens the scope. Apple says it found traces of systemic behavior: confidential documents photographed before interviews, discussions of unannounced products in private meetings between departing employees, company devices retained after resignation. The fact that multiple former employees contacted Apple to return devices after the complaint suggests the number of people involved could grow.

OpenAI published a sharp response. It says Apple got the wrong person in initial contact due to similar surnames, that Apple lied about a conversation with their general counsel, and that residual access to Apple systems by former employees stemmed from lackluster Apple security practices. On the merits, OpenAI states it does not have and does not want Apple trade secrets.

What this means for those building with AI. The case highlights a governance dimension that AI papers and playbooks rarely address: managing knowledge transfer when people change companies. In agentic systems, where employees instruct agents, craft prompts, curate datasets, and set tool boundaries, the departing person carries knowledge that now lives outside their head: in reusable prompts, in MCP configurations, in evaluation pipelines. Companies shipping agents to production are accumulating procedural IP that doesn’t sit in a policy document but in code, prompts, and configurations.

The limits of what we know. We’re in the preliminary motion phase: Apple makes accusations, OpenAI denies them, and the court hasn’t yet determined the facts. The detail on 11 additional employees comes from a filing, not from established fact. What can be said today is that this case will expose a practice (knowledge transfer between labs) that the industry has so far managed through non-compete agreements and NDAs, and that agentic systems make harder to track.