JoyIn Claims OpenAI Copied Its Robot Ideas

JoyIn, a robotics startup based in Suzhou and backed by Ant Group, has formally accused OpenAI of stealing its intellectual property and mimicking its digital presence. The allegations were detailed in an open letter posted on the company’s WeChat account by CEO Guo Renjie on September 10. Coinciding with these accusations, JoyIn announced the launch of Aether, a new model designed to control humanoid robots.

Guo expressed strong conviction that OpenAI had directly harvested JoyIn’s concepts without alteration. “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 wrote, as translated by CNBC. While CNBC reported on the letter, neither they nor The Next Web (TNW) could independently verify the claims. OpenAI did not immediately respond to requests for comment.

Allegations of Technical and Design Theft

The open letter outlines three primary areas where JoyIn believes its work was appropriated: technical architecture, conceptual frameworks, and website design. JoyIn states it presented its “extraterrestrial visitor” framework in Silicon Valley on August 16. Just weeks later, on September 6, OpenAI chief scientist Jakub Pachocki published an essay titled “An Alien Mind.” Guo points to similarities in discussions regarding recursive self-improvement and the use of AI to optimize computing power.

Regarding visual design, Guo compares the landing page for OpenAI’s GPT-6 Astra, launched on September 3, with JoyIn’s Aether website, which went live at the end of July. He describes the similarity as “pixel-level” copying, specifically citing particle effects as the core aesthetic element taken from JoyIn’s site.

Legal Action and Product Launch

In response to these perceived infringements, JoyIn’s legal team has initiated proceedings to file a lawsuit. The company also stated that it accelerated the release of the Aether model specifically because of OpenAI’s actions.

Pachocki’s essay addresses many of the themes raised by Guo. He argues that machine recursive self-improvement is central to future scientific discovery and essential for maintaining a competitive edge. He also notes that AI will enhance the computational substrate itself, a reference to OpenAI’s custom chip, Jalapeño. The concept of an alien intellect appears in both the title and the text of his essay. Furthermore, Pachocki mentions that hiding models’ chain-of-thought processes helps prevent distillation, the practice of training one model on the outputs of another.

Aether Model Specifications

JoyIn provides specific metrics for the Aether model, claiming it consists of 4 billion parameters. According to the company, the model was trained on 200 hours of human video data but used no data from actual robots. JoyIn asserts that Aether achieves a success rate above 90% on the first attempt without task-specific training. The system is designed to run on hardware from competitors like Unitree and AgiBot.

The architecture mimics brain regions ranging from the prefrontal cortex to the spinal cord. It features five core capabilities: active perception, multimodal understanding, streaming memory, motor dynamics, and self-iteration. Development was led by Zhu Mingxuan, who previously worked at Figure. Zhu told CNBC that Aether reduced training time by two-thirds. She indicated that while some components, such as touch sensitivity, would be open-sourced, energy-related parts would remain proprietary. TNW has not verified these performance figures.

Broader Context of Distillation Accusations

The dispute highlights broader tensions regarding AI data practices. US agencies, including the NSA, CISA, and FBI, recently issued an advisory accusing six Chinese firms,DeepSeek, Moonshot AI, Alibaba, MiniMax, StepFun, and Z. AI,of distilling American models like Claude, GPT, Gemini, and Grok. The advisory suggests this activity occurred with likely government awareness. China’s foreign ministry rejected these accusations. CNBC noted that JoyIn’s letter effectively flips the script, accusing OpenAI of the very practice that Washington has criticized Beijing for.