JoyIn founder Guo Renjie asserts that OpenAI’s recent model framework and design elements mirror JoyIn’s published ideas, such as pervasive intelligence networks and the recursive self-improvement concept, as well as AI-assisted optimization of computing power.

context.

CNBC could not independently verify JoyIn’s claims; while some concepts exist in AI research, implementations may differ and independent corroboration is needed.

The report notes irony in the broader industry context, where ownership disputes echo between Western firms and Chinese companies over model distillation and IP in AI development.

JoyIn, a Chinese humanoid robotics startup backed by Ant, accuses OpenAI of copying JoyIn’s key technical concepts and its Aether model, including ideas like recursive self-improvement and AI-optimized computing, and points to the space-themed design of OpenAI’s GPT-6 Astra webpage as part of the likeness.

CEO Guo Zhenjie frames the dispute in the context of an open letter and OpenAI’s September publication An Alien Mind, arguing similarities in recursive self-improvement concepts and AI-optimized computing capacity.

The broader AI safety and pace debate frames the story, with discussions about slowing development; OpenAI reportedly paused certain trainings, while leaders vary in optimism about rapid progress.

CNBC could not independently verify JoyIn’s claims; some concepts exist in AI research, but implementations differ among companies.

Anthropic has previously accused Chinese firms of distilling models, and a US cyber agency report lists six Chinese companies doing so; OpenAI had not commented at the time.

JoyIn’s strategy emphasizes a perception-driven paradigm and rapid training-time reductions within a competitive US-China tech landscape, noting legal disputes and public perception may influence collaboration and investment.

Anthropic has claimed multiple Chinese companies copied Claude, illustrating a reciprocal pattern of copying allegations across the AI field.