Skip to Content

Anthropic updates its usage policy: requirements for high-risk uses, autonomous robots and deceptive campaigns from 12 November

The new version of the Usage Policy takes effect on 12 November 2026 and spells out requirements for high-risk uses, autonomous hardware control, surveillance and unsupported regions.
October 8, 2026
Curated Medium

What it says

  • The updated policy takes effect on 12 November 2026.
  • It creates the section "Do Not Engage in Deceptive Campaigns or Artificial Activity", which groups the prohibitions on fake accounts and influence operations, whether political or commercial.
  • In high-risk uses (health, legal rights, finance, livelihoods, essential services) it maintains the requirement of a qualified human in the loop and of informing the person affected about the use of AI, and rewrites the section to list which recommendations are covered.
  • It adds requirements when Claude controls hardware with autonomous physical actions: a qualified operator must be able to observe and stop the equipment, which must maintain a safe state if Claude disconnects.
  • It prohibits using Claude to decide or recommend whom to investigate, detain or accuse, and clarifies that the regional ban extends to entities incorporated in those regions or majority-controlled from them.

The Tech Guardian summarises and comments in its own words on third-party publications. The full information is in the original source, which we recommend you read.

Our take

If you integrate Claude into processes that affect customers or employees, such as credit, recruitment or health, check before 12 November that you have a human reviewer with real authority and that you inform the person affected about the use of AI. The requirement is not new, but the list of covered cases is now explicit. Document that control: it will also serve you under the GDPR and the AI Act.

Ten agencies, including Spain's CNI, detail how the Chinese cyber-espionage network backed by Integrity Technology Group operates
Joint advisory AA26-281A describes how actors linked to the Chinese Government combine mass scanning, botnets and email theft on Exchange and Microsoft 365. It includes indicators and the exploited CVEs.