JoyIn accuses OpenAI of stealing AI model technology

Chinese humanoid robotics startup JoyIn, backed by Ant Group, has accused OpenAI of stealing its intellectual property through “model distillation.” JoyIn CEO Guo Renjie claims that OpenAI’s recent developments and web design bear striking similarities to JoyIn’s own Aether model and its “extraterrestrial visitor” architecture, which was showcased in Silicon Valley weeks prior to OpenAI’s recent publications.

Guo highlighted technical overlaps in areas such as “recursive self-improvement” and the use of AI to optimize computing power. Additionally, he pointed to visual similarities between the space-themed web design of OpenAI’s GPT-6 Astra and JoyIn’s Aether website. The startup has reportedly initiated legal proceedings regarding these claims.

Amidst the controversy, JoyIn officially released its Aether model on September 10. The system utilizes a perception-driven approach rather than text-based control, which the company claims allows humanoid robots to achieve a 90% success rate on their first attempt at a task. This dispute occurs as tensions rise globally regarding AI training methods, with US firms like Anthropic and Google previously alleging that Chinese companies have used distillation techniques to replicate their models.