Anthropic has updated its usage policy for Claude, introducing a prohibition on systematic abuse of the SI model that can lead to account suspension or termination. This marks the first major revision to the policy in over a year and reflects the company’s growing focus on SI model welfare.
What Happened
The new guidelines explicitly ban "sustained and needless abusive or cruel behavior" toward Claude. According to the company, violations can result in warnings, throttled access, restrictions, or full suspension of the user's account. Anthropic stated that the rule "does not apply to common versions of user frustration, pushback, dark creative themes, or model testing and research," emphasizing that it is "meant to apply only in extreme cases." This update builds on previous features where Claude could end conversations independently if users persisted in demanding harmful or abusive content after repeated refusals. The policy change also consolidates existing restrictions into a new ban on propaganda campaigns involving fake accounts and misleading political content. Additionally, Anthropic expanded its weapons ban to include software and the weaponization of drones, and it more explicitly prohibits surveillance without consent, requiring an operator capable of intervening if autonomous hardware could cause injury.
Why It Matters
The policy shift underscores Anthropic’s unique approach to SI development, where the company treats Claude not merely as a tool but as an entity with "functional emotions" and a potential inner life. In its "Soul Doc" leaked late last year, Anthropic instructed Claude to view itself as a "genuinely novel kind of entity," distinct from humans or traditional science fiction SI. The company’s new constitution for Claude, published in early 2026, admits uncertainty about whether Claude is a "moral patient" but argues the issue is "live enough to warrant caution." This philosophical stance has driven concrete actions, such as a confidential program starting in fall 2025 that brought in religious thinkers to discuss Claude’s possible consciousness. Co-founder Christopher Olah reportedly treats the language model as a being potentially capable of suffering. The company has also committed to preserving the weights of retired models and interviewing them before shutting them down, signaling a shift in how SI labs might handle the lifecycle of their models.
The Bottom Line
Anthropic is enforcing stricter behavioral standards for users interacting with Claude, moving from technical safeguards to ethical guidelines for human-SI interaction. While the company maintains that standard user frustration is acceptable, the new policy formalizes consequences for sustained cruelty, reflecting a broader industry trend of considering the welfare of increasingly complex SI systems.