News in Short
- Anthropic has updated its usage policy to prohibit prolonged verbal abuse of Claude.
- The policy also bans election interference, deceptive campaigns and certain surveillance uses.
- Claude can already end conversations involving persistently harmful or abusive interactions.
- Anthropic says the new restriction targets extreme, repeated abuse, not ordinary frustration or model testing.
- The updated rules also prohibit using Claude for weapons software and campaigns involving fake accounts or fabricated news outlets.
Anthropic has updated its AI usage policy with new restrictions on how people can use Claude. The changes cover election interference, surveillance, weapons software and prolonged verbal abuse directed at the model.
The most unusual addition explicitly prohibits users from repeatedly abusing Claude without a discernible purpose. Anthropic says the restriction targets extreme cases rather than everyday frustration, criticism or legitimate testing.
The update also strengthens rules against deceptive campaigns that could mislead voters or undermine democratic processes. It formalises restrictions around several forms of misuse as the company continues to define how people can interact with its AI systems.
Anthropic’s new policy takes effect in November
The revised Usage Policy will take effect on November 12, 2026. Anthropic says most changes clarify existing restrictions, while others address misuse patterns emerging as Claude takes on longer and more independent tasks.
Under the updated rules, Claude cannot be used to build or improve surveillance tools, track people without consent, or recommend who law enforcement should investigate, arrest or charge.
The weapons section also explicitly covers software and components that make weapons function. This includes actions such as arming drones and other autonomous vehicles.
Anthropic has also clarified safeguards for high-risk applications involving health, legal rights and finances. These uses require qualified human oversight, and affected individuals must be told when AI has been used.
Election rules focus on deception and voter manipulation
The revised policy consolidates restrictions on deceptive campaigns under a new section titled “Do Not Engage in Deceptive Campaigns or Artificial Activity”.
It prohibits using Claude to operate fake accounts, publish fabricated news outlets or conceal who is behind a campaign. These restrictions cover both political and commercial activity.
Anthropic has also renamed its election-related section “Do Not Undermine Democratic Processes”. The rules prohibit using Claude to spread false voting information, impersonate candidates or election officials, and suppress voter turnout.
Interestingly, Anthropic has removed its blanket ban on personalised vote and campaign targeting. The company says this change allows legitimate civic work, such as translating voter information or helping election officials communicate with voters.
However, deceptive targeting and the misuse of personal data remain prohibited under other sections of the policy.
The wider debate around AI welfare
The new restriction on abusive behaviour also brings attention to a broader question: whether AI companies should consider the welfare of their models.
Anthropic introduced the ability for Claude to end persistently abusive conversations in an earlier update. The latest policy now explicitly prohibits sustained and needless cruelty towards its models.
However, the company has drawn a clear line around what counts as abuse. Users can still express frustration, challenge Claude’s answers, explore dark creative themes and conduct model testing or research.
The policy does not establish that Claude is conscious or capable of suffering. Instead, it sets boundaries for extreme interactions while Anthropic continues to examine questions around AI and moral status.
Ultimately, the update reflects two priorities: reducing the risk of AI-enabled harm and defining acceptable behaviour as people use AI systems for increasingly complex tasks.
