OpenAI Faces JoyIn Distillation Claim

OpenAI Faces Distillation Claim From Chinese Humanoid Robotics Startup

Startups

OpenAI faces a fresh distillation claim from Chinese AI startup JoyIn as its Aether robotics model raises questions over technology and design similarities.

OpenAI faces a direct distillation claim from Chinese robotics startup JoyIn.
JoyIn points to striking similarities in AI concepts and website design.
A lawsuit may be coming, while OpenAI has yet to respond publicly.
JoyIn Aether targets humanoid robots, claiming 90% first-attempt task success.
The clash puts AI model ownership under fresh scrutiny as U.S. and Chinese firms push deeper into robotics.

OpenAI is facing a fresh accusation from a Chinese humanoid robotics startup, after JoyIn CEO Guo Renjie published an open letter questioning similarities between the companies’ AI work.

Guo, who leads the Ant-backed startup based in Suzhou, said JoyIn had presented its “extraterrestrial visitor” AI framework in Silicon Valley weeks before OpenAI Chief Scientist published “An Alien Mind” on Sept. 6.

He pointed to similar ideas, including recursive self-improvement and using AI to improve computing efficiency. He also questioned similarities between OpenAI’s GPT-6 Astra website and JoyIn’s Aether website, which he said had been released two months earlier.

But is this really a case of copying, or two companies reaching similar ideas?

That question is central to the dispute. Some of the concepts cited by Guo are already part of wider AI research, even though companies may use them in different ways.

OpenAI had not immediately responded to a request for comment.

Guo went further, saying JoyIn had started the process of filing a lawsuit. His accusation also carries an unusual twist because U.S. AI companies have repeatedly accused Chinese firms of using AI Model Distillation to reproduce the capabilities of their systems.

Could the distillation debate now turn back on the companies making those accusations?

JoyIn used the moment to introduce its JoyIn Aether model publicly. The startup says its perception-based approach to robotic control allows humanoid robots to complete tasks with a 90% first-attempt success rate.

Zhu Mingxuan, who led Aether’s development, previously worked at U.S. humanoid robotics company Figure. She said Aether cut training time by two-thirds and that JoyIn plans to open-source parts of the system, including technology related to touch.

The dispute comes as Chinese AI startups and U.S. technology firms pursue increasingly capable AI and humanoid systems.

Who has the stronger case may depend on what can actually be proved.

Business Honor examines that JoyIn has made a serious claim, but not a proven one. The bigger issue is whether AI distillation standards can be applied consistently when similar research ideas appear across competing companies.