Anthropic updates usage policy to ban sustained mistreatment of Claude

Anthropic updated its usage policy on October 8, 2026, to prohibit users from subjecting its Claude artificial intelligence model to sustained and needless abusive or cruel behavior. MacRumors and The Verge both reported on the policy change, which marks the first update to the guidelines in over a year. The Telegraph, Forbes, Yahoo Tech, Newser, and Firstpost also reported on the restriction. [1][2][5][6][7][8][9]

According to MacRumors, Anthropic stated that the rule is intended for extreme cases where users repeatedly act cruelly toward the artificial intelligence models with no discernible purpose. The Verge noted that the ban does not apply to normal user frustration, pushback, dark creative themes, or modified usage. Last August, Anthropic announced that Claude could end conversations with persistently harmful or abusive users. [1][2]

The Verge reported that the updated policy addresses broader high-risk misuse cases for the first time in over a year. The new guidelines include restrictions covering election interference, weapons development, surveillance, propaganda campaigns, as well as health and financial uses. [2]

The policy update follows recent product developments from the company. SiliconANGLE reported that Anthropic released its Claude Haiku 5.5 small model, pricing it at roughly one quarter of the cost to run Haiku 4.5. Android Authority and SiliconANGLE noted that the launch came two weeks after Opus 5.5 was released on September 22. SiliconANGLE added that Anthropic halved the price charged for cache reads on Claude Sonnet 5.5. [3][4]