The Claude policy update adds autonomous-hardware safeguards and clarifies rules for high-risk decisions, unsupported regions, weapons and surveillance.
Anthropic’s revised Usage Policy for Claude takes effect on November 12, 2026. Published on October 8, the update responds to models undertaking longer, more independent work and to misuse patterns that the company says it observed in influence operations, weapons development and surveillance. Anthropic describes most of the revisions as clarifications of existing rules, alongside new controls for autonomous physical actions and abusive behavior toward models.
A consolidated section now prohibits deceptive campaigns and artificial activity in political and commercial settings. It covers concealing who is behind a message, amplifying material through fake accounts or posts, and building tools or infrastructure for influence operations. Anthropic says these restrictions previously appeared across its election, fraud, privacy and disinformation rules.
The elections section is now focused on attempts to deceive voters or disrupt elections, including false information about candidates or voting, impersonation of candidates or election officials, and turnout suppression. Anthropic has removed its blanket ban on personalized voter and campaign targeting because that rule also covered legitimate civic work, such as multilingual voter information and ballot-cure notices. Targeting based on deception or misuse of personal data remains prohibited elsewhere in the policy.
Weapons restrictions now expressly cover guidance and control software, components that make weapons function, and actions such as arming drones or other autonomous vehicles. The surveillance provisions prohibit non-consensual tracking, including analysis of previously collected data, and bar Claude from deciding or recommending whom authorities should investigate, arrest or charge. They also prohibit building or improving tools designed for surveillance. Anthropic says these formulations make its existing enforcement more explicit; consensual fraud monitoring, content moderation, journalism and legal research remain permitted examples.
For uses affecting health, legal rights, finances, livelihoods or access to essential services, a qualified person must be able to review and, when necessary, change Claude’s recommendation. The affected individual must also be told that AI was used. Anthropic stresses that these requirements have not changed: the rewritten section more directly identifies which recommendations are covered and which are not.
New safeguards apply when Claude is connected to hardware that takes autonomous physical actions and could cause injury. A qualified operator must be able to observe and stop the equipment, which must retain a safe state if disconnected from Claude. The policy also newly prohibits sustained, needless cruelty toward models in extreme cases without a discernible purpose, while excluding ordinary frustration, pushback, dark creative themes, testing and research.
Anthropic also clarified its Supported Regions Policy. Claude may not be used by people physically located in unsupported regions; entities incorporated or headquartered in those regions; or entities majority-owned or controlled by people or organizations there. This is broader than a location check on an individual user because incorporation, headquarters, ownership and control can each determine whether an organization is eligible to use Claude.
Organizations preparing for November 12 therefore need to review both Claude workflows and surrounding product controls. The policy distinctions are material: human review and AI disclosure for high-risk recommendations are existing requirements, while autonomous equipment receives added operator-observation, stopping and safe-state safeguards. Eligibility checks must also account for physical presence and corporate ownership or control under the Supported Regions Policy.
Sources
Originally published by AI Feed: Anthropic’s Revised Claude Usage Policy Takes Effect November 12.
Top comments (0)