Chinese firms used Anthropic's AI model, violated US export rules
Anthropic, a US AI startup, claims three Chinese companies—DeepSeek, Moonshot AI, and MiniMax—illicitly used its Claude AI model.
They allegedly used 24,000 fake accounts to interact with Claude and create their own versions, violating Anthropic's terms of service and its regional access restrictions.
Distilled models could be misused for surveillance, cyber threats
This isn't just tech drama—it's about global security.
The companies used a process called "distillation" to make smaller, cheaper versions of the AI that are easier to copy and use.
Anthropic warns these distilled models lose important US safety protections and could be misused for things like surveillance or cyber threats.
With recent US-China tech tensions rising (especially after the NVIDIA chip export approval), Anthropic is urging more teamwork between tech companies and policymakers to keep things safe.