Realtime AI News
Kimi K3 Brought Moonshot AI About $1B in Revenue, While Anthropic Claims Claude Was Used Covertly
According to Incrypted, Moonshot AI's Kimi K3 brought the Chinese company roughly $1 billion in revenue, a sign of how large a commercial footprint a Chinese frontier model can build. The same report says Anthropic claimed Claude was used covertly, turning attention to questions about training processes and data provenance.
Moonshot AI's Kimi K3 brought the Chinese company about $1 billion in revenue, while Anthropic claimed that Claude was used covertly, according to a report from Incrypted. The story places a commercial milestone and an allegation about training compliance side by side.
The revenue figure is the most concrete element. A Chinese frontier model being credited with billion-dollar-scale revenue suggests the Kimi line has built a real paying base rather than remaining a technical showcase.
The more contentious part is Anthropic's claim. As reported, Claude is said to have been used in a non-public way, an allegation that typically touches on model distillation, the provenance of training data and compliance with service terms, and one that sits inside the longer-running technology and business rivalry between US and Chinese AI companies.
At this stage the report relays competing claims rather than a settled verdict. Until independent technical evidence, a formal company response or a regulatory statement appears, the dispute is likely to remain a matter of assertion on both sides.
Why it matters: if leading labs begin trading public accusations about training compliance, the fallout reaches beyond a single company to the developers and enterprises that rely on cross-border model services, APIs and distillation, whose compliance costs and tooling choices could shift.
What to watch next: whether Moonshot AI responds, whether Anthropic supplies more technical detail, and whether the dispute moves from public argument into legal or policy channels.
Why it matters
If the training-compliance dispute escalates into open confrontation, the boundaries for cross-border model services and distillation could tighten, making both companies' responses worth tracking.
Nearby Updates
All09/13, 18:59
Zhipu Raises About $5 Billion to Fund Next-Gen GLM, Self-Training and Compute
Zhipu AI (02513.HK) said on September 13 that it completed roughly $5 billion in financing, split between about $2 billion in share placement and about $3 billion in zero-interest convertible bonds. Proceeds will go to its next-generation GLM foundation models, a fully self-training system and related compute infrastructure.
09/13, 17:30
Black Lake founder pitted an AI sales coach against his most experienced leader — 7 of 10 senior salespeople chose the AI
Black Lake Technologies founder Yuxiang Zhou pitted an AI sales coach against his company's most experienced sales leader, and seven of the startup's 10 most senior salespeople voted for the machine's advice. The Shanghai company, which sells AI agents to factories across Asia, fed recorded sales conversations into an agent trained to coach, and Zhou says he is still waiting for AI's 'Google moment.'
09/13, 15:50
NASA and IBM Release Open-Source Lunar Foundation Model Trained on 30 Data Layers
NASA and IBM have released an open-source Lunar Foundation Model that fuses more than 30 layers of observations from nine instruments across four lunar missions into a single multimodal system for crater detection, volcanic mapping and ice prospecting. Against the SwinV2 baseline it cut RMSE on ice prospectivity by 22%, improved IoU on irregular mare patches by 3%, and raised accuracy and mAP on 100-metre crater detection by 19% while using half the training data.
09/13, 15:38
One Navigation Model, Four Robot Bodies: LightNav-0 Turns 2,000+ Real Scenes Into Simulation
Liangyuan has unveiled LightNav-0, a general navigation model whose Real2Sim2Real engine converts more than 2,000 internet-sourced real scenes into reusable simulation environments and yields over 4,000 hours of vision, language and action post-training data. The same model is deployed zero-shot on humanoid, quadruped, wheeled and flying robots, with model, code and a technical report released alongside it.