Field note
Apple filed a federal trade-secrets case against OpenAI, io Products, Chang Liu, and Tang Tan on July 10. Apple alleges that former employees brought confidential hardware information into OpenAI's device effort and is seeking a preliminary injunction and expedited discovery. These are allegations; the court has not ruled on them.
OpenAI responded publicly on August 3, disputing Apple's factual account, saying Apple confused identities in parts of its outreach, and stating that it does not want Apple's trade secrets. OpenAI also published communications around the dispute. The competing narratives are now part of the evidence fight, not a settled history.
The broader engineering question is concrete. Hardware and agent-device programs recruit people precisely for accumulated expertise, while employment agreements protect particular documents, prototypes, designs, and non-public processes. A defensible organization needs clean-room boundaries, provenance for imported design decisions, prompt access revocation, and written review when a new hire proposes something close to a former employer's system.
The news is the escalation into litigation and OpenAI's unusually public rebuttal. The durable lesson is that talent mobility becomes an evidence-management problem when teams are building new physical interfaces for AI.
