Overview
- A joint advisory from the NSA, CISA and the FBI on Tuesday named DeepSeek, Moonshot AI, Alibaba, MiniMax, StepFun and Z.AI and accused them of large-scale 'knowledge distillation' against U.S. frontier models.
- The agencies said the operations ran since late 2024, used millions of requests to extract billions of tokens from models such as Anthropic’s Claude, OpenAI’s GPT/ChatGPT, Google’s Gemini and xAI’s Grok.
- Officials described sophisticated evasion tactics including routing queries across many accounts, APIs, cloud providers, aggregators and proxies to hide metadata and bypass usage controls.
- Beijing has formally rejected the charges and warned of countermeasures, while U.S. officials say the advisory raises the prospect of sanctions, Entity List designations and export or legislative steps before the Trump‑Xi summit.
- The advisory urges U.S. firms to adopt detection and mitigation measures, share intelligence and consider output‑level defenses, noting that preventing covert distillation is technically hard and may affect access for legitimate users.