A Chinese humanoid robotics startup backed by Ant Group has publicly accused OpenAI of copying its technology, escalating a dispute over model distillation into a rare reversal of the usual direction of such claims.
Guo Renjie, CEO of Suzhou-based JoyIn, published a letter to OpenAI in Chinese on Thursday alleging that the U.S. company’s recent work closely tracks his startup’s own models and design language. “People often say major tech companies have intelligence networks monitoring the whole internet, this time I believe it, this is a direct distillation of us without any modifications,” Guo said in the statement, according to a CNBC translation. He added that JoyIn has started the process of filing a lawsuit.
What JoyIn Claims OpenAI Copied
Guo said JoyIn publicly presented its “extraterrestrial visitor” AI model framework in Silicon Valley a few weeks ago, before OpenAI’s chief scientist published a paper titled “An Alien Mind” on Sept. 6. He pointed to similarities in core technical approaches, including “recursive self-improvement” and the use of AI to optimize computing power.
He also flagged the outer space-inspired design of OpenAI’s GPT-6 Astra web page, which he said resembles the site for JoyIn’s Aether model that he claims was released two months ago. CNBC said it was unable to independently verify the claims, and noted that some of the concepts involved are established parts of AI research more broadly, even as companies connect and implement them differently. OpenAI did not immediately respond to a request for comment.
Aether’s Pitch: Perception Over Text
JoyIn used the dispute to draw attention to its own model. Guo said he decided to publicly announce the Aether model on Thursday. According to the company, Aether takes a perceptive rather than text-based approach to robotic control, allowing humanoids to complete tasks with a 90% success rate on the first attempt.
Zhu Mingxuan, who led the model’s development, said she left U.S. humanoid company Figure last year, where she had also worked on models for helping humanoid robots mimic human actions. She told CNBC earlier this week that Aether reduced training time by two-thirds and that she planned to open-source parts of the model, such as those relating to touch, but not portions related to energy use.
Distillation Claims Usually Run the Other Way
The accusation inverts a pattern that has become familiar in U.S.-China AI competition. U.S.-based Anthropic has repeatedly flagged unauthorized “distillation” of its models by Chinese companies seeking to improve their own capabilities. On Tuesday, a U.S. cybersecurity agency said six Chinese companies, including DeepSeek and Alibaba, distilled models from Anthropic, Google and OpenAI.
Distillation generally refers to training a smaller or newer model on the outputs of an existing one, a practice that sits in a gray area of AI development.
Not everyone in the U.S. startup world treats the practice as a serious threat. Garry Tan, chief executive of startup accelerator Y Combinator, told CNBC this month that he would “do nothing” about distillation, joining others who have pointed out that U.S. companies trained their own AI models on data covered under copyright law.
Behind the sparring over attribution, the harder engineering problem remains: getting humanoid robots and the AI models that guide them to perform as intelligently as humans across a wide range of tasks. That challenge sits at the center of a broader technology race between U.S. and Chinese companies, one that JoyIn’s public letter has now drawn into a dispute over who copied whom.
CNBC’s Kate Rooney and Isabel O’Brien contributed to this report.
Source: www.cnbc.com — https://www.cnbc.com/2026/09/11/chinese-humanoid-robot-startup-distillation-claim-openai.html
This article is for informational purposes only and does not constitute financial, investment, tax, or legal advice. Do your own research and consult a licensed professional before making financial decisions.
