Anthropic Updates Claude Usage Policy, Adding Ban on Model Abuse and Other Terms

On October 9, 2026, The Verge reported that Anthropic updated Claude's usage policy for the first time in over a year, adding a clause prohibiting users from engaging in sustained and gratuitous cruelty or abuse toward the model. The report noted that ending conversations remains the primary enforcement mechanism, and the clause applies only to extreme cases of purposeless, repeated abuse—not to general negative feedback or heated conversations. TechCrunch subsequently reported additional details: the update also added explicit bans on election interference, weapon software, and surveillance; the policy clarifies that the bans do not cover users' common frustration, dark creative themes, or model testing research; and a separate section prohibits using Claude to operate fake accounts, fictitious news outlets, or mislead voters. The Decoder later reported that violators of the model abuse rule can be warned, rate-limited, restricted, or have access terminated—more specific than the earlier claim that 'ending conversations is the primary enforcement mechanism'; the report also noted the new policy consolidated restrictions on fake-account propaganda, weaponized drones, and non-consensual surveillance, and acknowledged that government contracts may be exceptions. The update has drawn outside attention to the topic of AI model welfare.

AI overview of reporting · Updated 10 days ago