The CEO of an Ant-backed humanoid robotics startup has published a letter to OpenAI in Chinese that raises questions about technical and design similarities between the two companies' recent models.
"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 Renjie, CEO of Suzhou-based JoyIn, said in a public statement Thursday, according to a CNBC translation.
He said the startup had publicly presented its "extraterrestrial visitor" AI model framework in Silicon Valley a few weeks ago, before OpenAI's Chief Scientist published "An Alien Mind" on Sept. 6. Guo questioned similarities in core technological approaches such as "recursive self-improvement" and the use of AI to optimize computing power.
Guo also highlighted similarities in the outer space-inspired design of OpenAI's GPT-6 Astra web page and the website for JoyIn's Aether model that he claimed was released two months ago. He added the startup has started the process of filing a lawsuit.
CNBC was unable to independently verify the claims. Some concepts are existing parts of AI research more broadly, although companies connect and implement them differently. OpenAI did not immediately respond to a request for comment.
U.S.-based Anthropic has repeatedly flagged unauthorized "distillation" of its models by Chinese companies to improve their own AI capabilities. On Tuesday, a U.S. cybersecurity agency said six Chinese companies, including DeepSeek and Alibaba, distilled models from Anthropic, Google and OpenAI.
JoyIn's Aether model, which Guo said he decided to publicly announce Thursday, claims its perceptive, rather than text-based, approach to robotic control enables humanoids to complete tasks with a 90% success rate on 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.
Getting humanoids and AI models to perform as intelligently as humans across a variety of tasks has remained a challenge, amid a broader tech race between U.S. and Chinese companies.
Garry Tan, chief executive of famed startup accelerator Y Combinator, told CNBC this month that he would "do nothing" about distillation and joined others in pointing out that the U.S. companies have trained their AI models on data covered under copyright law.
—CNBC's Kate Rooney and Isabel O'Brien contributed to this report.
Facts Only
* Guo Renjie is the CEO of JoyIn, a Suzhou-based humanoid robotics startup.
* JoyIn is backed by Ant.
* OpenAI's Chief Scientist published "An Alien Mind" on September 6.
* Guo Renjie published a letter to OpenAI in Chinese on Thursday.
* JoyIn presented its "extraterrestrial visitor" AI model framework in Silicon Valley several weeks prior to September 6.
* JoyIn released the Aether model website two months ago.
* OpenAI's GPT-6 Astra web page features an outer space-inspired design.
* JoyIn has initiated the process of filing a lawsuit.
* A U.S. cybersecurity agency reported on Tuesday that six Chinese companies, including DeepSeek and Alibaba, distilled models from Anthropic, Google, and OpenAI.
* Zhu Mingxuan, lead developer of Aether, previously worked at Figure.
* JoyIn claims the Aether model achieves a 90% success rate on first attempts for humanoid tasks.
* Garry Tan is the CEO of Y Combinator.
Executive Summary
Guo Renjie, CEO of the Ant-backed startup JoyIn, has accused OpenAI of "distilling" JoyIn's proprietary AI frameworks and design elements. The dispute centers on technical similarities regarding recursive self-improvement and computing power optimization, as well as aesthetic parallels between the GPT-6 Astra web page and JoyIn's Aether model site. JoyIn asserts these similarities emerged after their framework was presented in Silicon Valley, and the company has begun the process of filing a lawsuit.
The situation unfolds against a broader geopolitical backdrop of AI competition. While U.S. agencies and companies like Anthropic have flagged unauthorized distillation by Chinese firms to accelerate their capabilities, some U.S. industry leaders, such as Y Combinator's Garry Tan, express indifference toward distillation, citing the widespread use of copyrighted data in training. Whether JoyIn's claims represent a genuine theft of intellectual property or the simultaneous emergence of common AI research trends remains unverified, as OpenAI has not commented on the allegations.
Full Take
The strongest version of this narrative is that a smaller, agile innovator has had its core breakthroughs absorbed by a dominant industry titan through "distillation"—the process of using a larger model to mimic the behavior of another—essentially automating intellectual property theft.
This situation is a textbook study in the "Gray Zone" of AI development. The load-bearing element of the claim relies entirely on temporal correlation (presentation in Silicon Valley followed by a release) and aesthetic similarity. Because the technical "recursive self-improvement" mentioned is a known pursuit in broad AI research, the narrative teeters between a specific accusation of theft and a general observation of convergent evolution in tech.
Patterns detected: none
The driving paradigm is the "AI Arms Race," where the distinction between inspiration, distillation, and theft is blurring. This echoes historical industrial espionage patterns but at the speed of software. The unstated assumption is that "distillation" is a zero-sum game; however, if the underlying concepts are indeed existing parts of broader research, the claim of theft becomes a strategic move for visibility or legal leverage.
The implications for human agency are significant: if the dominant players can simply "distill" the innovations of smaller startups, the incentive for independent, high-risk research may diminish, leading to an oligopoly of intelligence. Conversely, if "distillation" is accepted as standard practice, the concept of intellectual property in the age of weights and biases may become obsolete.
Bridge Questions:
1. How can a third party mathematically distinguish between convergent research and model distillation?
2. If distillation becomes the industry norm, what new frameworks for intellectual property must emerge to protect smaller innovators?
3. To what extent is this legal threat a marketing strategy to signal the viability of the Aether model?
Counterstrike Scan: A coordinated campaign would weaponize this to frame U.S. tech giants as predatory "imperialists" to galvanize nationalist support for domestic Chinese AI. The current content does not match this; it remains a business-centric report on a specific legal dispute.