GoKawiilUS agencies accuse six Chinese AI firms of mass model-distillation attacks on US chatbots
CISA, the NSA, and the FBI issued a joint advisory naming DeepSeek, Moonshot AI, Alibaba, MiniMax, StepFun, and Z.AI as having harvested billions of tokens from Anthropic, OpenAI, Google, and xAI models since late 2024. The firms allegedly used fraudulent accounts, proxy networks, and automated failover systems to bypass rate limits and extract restricted reasoning data at industrial scale.
GoKawiil's interpretation of the reporting above, not reported fact.
If accurate, this shows how Chinese AI labs may be shortcutting years of costly training by siphoning outputs from leading US models rather than building comparable systems from scratch. It also signals that Washington now views this practice as state-aware strategic behavior, not isolated corporate misconduct, potentially inviting new export controls or API access restrictions.
- Six named Chinese AI firms allegedly extracted billions of tokens from major US frontier models since late 2024.
- Techniques included proxy networks, fraudulent accounts, and attempts to extract hidden chain-of-thought reasoning.
- US agencies say the scale suggests government awareness and a deliberate strategy to cut AI development costs and time.
Yubico YubiKey 5C NFC Security Key — When the story is about fraudulent and shared API accounts being abused at scale, hardware-backed account security suddenly feels very concrete. The YubiKey 5C NFC adds phishing-resistant two-factor login to accounts like Google, GitHub and OpenAI, tapping or plugging in instead of relying on codes anyone can steal. It's a pocket-sized upgrade for developers who hold keys to expensive AI services.
See Yubico YubiKey 5C NFC Security Key on Amazon → Affiliate link — we may earn a commission on purchases, at no extra cost to you. Product picked by AI based on this article; it is not a tested recommendation.Source: bleepingcomputer.com, 2026-09-09
Published there as: “US says Chinese firms extracted billions of tokens from frontier AI models”
Read the original report → The summary and analysis above are GoKawiil's own, written from reporting by the source above. Facts and quotes belong to the original publisher.