The NSA, the FBI and CISA are advising US AI companies to discreetly distort the responses sent to users suspected of attempting to train a rival model

Six Chinese laboratories are being targeted, including DeepSeek and Alibaba, accused of having significantly increased the number of queries sent to Claude, GPT, Gemini and Grok since late 2024 via intermediaries masking the origin of the requests

Anthropic had already reported in February that more than 16 million exchanges had been obtained via around 24,000 fraudulent accounts, but presenting ‘distillation’ as the basis for Chinese development goes beyond mere suspicious accounts