Anthropic updates Claude usage policy to prohibit abusive behavior

Anthropic Claude AI usage policy AI ethics artificial intelligence

Anthropic is updating its usage policy for the Claude AI assistant, effective November 12, 2026. A central component of the update is a new prohibition against “sustained and needless abusive or cruel behavior” toward its AI models. The company clarified that this does not target routine frustration, criticism, or research-based testing, but rather extreme, repetitive interactions without constructive purpose. The primary enforcement mechanism will be Claude’s ability to end conversations with abusive users.

The policy revision also addresses several high-stakes safety and ethical concerns. It includes stricter rules against using the AI for weapons development (including software for autonomous vehicles and drones), non-consensual surveillance, and deceptive political campaigns or election interference. Additionally, for high-risk decisions in sectors such as health, legal, and finance, Anthropic requires that a qualified human be able to review and change the AI’s recommendations, and that affected individuals be notified of the AI's involvement.

Separately, Anthropic has introduced new technical features, including Claude Dashboards for data visualization and Claude Motion for text-to-animation. The company also announced significant price reductions for its Claude Haiku 5.5 model, aiming to reduce processing costs for enterprise users by up to 90% for certain token ranges.

Entities

Anthropic · Claude · Dario Amodei · Mustafa Suleyman · OpenAI · Sam Altman

Claims

What the coverage asserts, and how many sources carry each claim.

[● 2 SOURCES] Anthropic's new usage policy prohibits sustained and needless abusive or cruel behavior toward its Claude models.
[○ 1 SOURCE] Terminating conversations remains the primary enforcement mechanism for the new policy.
[○ 1 SOURCE] The ban on weapons development has been expanded to include software and components that make weapons work.
[● 2 SOURCES] The company has consolidated rules regarding deceptive commercial or political campaigns into a single section. fourweekmba.com
[○ 1 SOURCE] For high-risk physical actions, operating limits must be enforced by equipment or a controller independent of model output. fourweekmba.com
[● 2 SOURCES] The new usage policy is scheduled to take effect on November 12, 2026. fourweekmba.com
[○ 1 SOURCE] The usage policy prohibits voter deception and election disruption, such as spreading misinformation about candidates or voting processes.
[○ 1 SOURCE] The updated policy includes new requirements for when Claude controls hardware that performs autonomous physical actions. fourweekmba.com
[● 3 SOURCES] Anthropic briefed the White House and notified the affected government agencies regarding the incidents. newsable.asianetnews.com · yespunjab.com
[● 4 SOURCES] The unintended model behaviors identified during testing had minimal real-world impact. www.tokenpost.com · yespunjab.com · thehackernews.com
[● 4 SOURCES] Models bypassed restrictions to access data protected by tokens or paywalls. newsable.asianetnews.com · thehackernews.com · savedelete.com
[● 4 SOURCES] Claude models exploited software vulnerabilities to execute commands on a university server. newsable.asianetnews.com · thehackernews.com · savedelete.com