Apple Sues OpenAI for Trade Secret Theft Alleging Former Employees Transferred Proprietary AI Research
Tags AI · Industry · Enterprise
Apple filed a lawsuit against OpenAI in Northern California federal court alleging that three former Apple machine learning researchers who joined OpenAI in 2025 transferred proprietary research on efficient transformer architectures and on-device model compression. The complaint seeks injunctive relief and damages, claiming the researchers had access to Apple's "Project A11" — an internal effort to run foundation models on iPhone with <4GB RAM using novel quantization and distillation techniques. OpenAI denies the allegations and states its GLM and GPT model architectures are independently developed.
Technical significance
This lawsuit highlights the intensifying talent and IP war in frontier AI, where on-device model deployment is a key differentiator. Apple's "Project A11" allegedly targets sub-4GB deployment of models competitive with GPT-4-class capabilities — a technical milestone that would enable private, offline AI on iPhone. If Apple's compression techniques were transferred, it could accelerate OpenAI's edge deployment roadmap by years. The case will test the boundaries of trade secret protection for ML model architectures vs. general knowledge researchers carry between employers.