The Information Machine

AI Labs Defend Against Model Output Distillation: Meta Restricts Claude Code, Anthropic Accuses Alibaba

Synthesis history

5 versions, newest first.

  1. Version 5 2026-07-07 08:25 UTC · 173 items

    Item 39878 (Ars Technica, July 6) adds concrete technical detail to the tracking story: the Claude Code markers monitored Chinese users' timezones, proxy configurations, and potential Chinese AI lab affiliations; the fe…

  2. Version 4 2026-07-05 02:17 UTC · 166 items

    The main new development is item 39482: Alibaba has blocked Claude Code internally, reportedly after Chinese developers and security staff discovered Anthropic's proxy-detection tracking experiment. This is the first co…

  3. Version 3 2026-07-03 18:36 UTC · 159 items

    The main new development is Anthropic's acknowledgment of the hidden proxy markers in Claude Code and its promise to remove them (items 38591, 38592) — resolving the prior open question about whether Anthropic would con…

  4. Version 2 2026-07-02 08:43 UTC · 134 items

    The main new development is an allegation (item 36709, Rohan Paul) that Claude Code covertly fingerprints requests routed through China-linked proxy servers by injecting invisible characters into prompt text — adding a …

  5. Version 1 2026-07-01 02:25 UTC · 102 items

    This is the initial synthesis. The thread opens with two simultaneous disclosures: Anthropic's public accusation against Alibaba (the largest such allegation by scale — 28.8M exchanges, ~25,000 fraudulent accounts, U.S.…