BABA
New AI beef, same old wall of allegations
Anthropic dropped a report saying several China-based AI companies may have been using Claude outputs in ways that look a lot less like normal benchmarking and a lot more like industrial-scale model distillation. Alibaba was the biggest name in the pile-up, with Anthropic claiming it detected more than 151 million exchanges tied to the company between May and July.
Why investors should care
If those allegations hold water, this isn’t just nerdy model-jockeying. It could mean:
- more legal and compliance headaches for Alibaba's AI push
- extra scrutiny around data sourcing and privacy practices
- a fresh reminder that the AI race is also a trench war over IP
The plot gets messier
Anthropic also named Moonshot and DeepSeek in the report, saying their activity included routing some requests to Claude and, in DeepSeek’s case, a large number of alleged distillation attacks. Daniel Newman then basically lit the fuse on X, arguing that some Chinese “open weight” models may owe more to U.S. frontier labs than people want to admit.
Big picture: AI is moving fast, but the legal bill is showing up right behind it.