Anthropic bans sustained abuse of Claude under new TOS

Anthropic's updated policy is its first revision in over a year and folds several existing restrictions into a clearer framework. The abuse rule builds on earlier models that could already end conversations when users kept demanding harmful content, a feature that grew out of Anthropic's research into AI model welfare. That research has led the company to treat Claude as more than a tool: in the leaked "Soul Doc," Claude is told to view itself as a "genuinely novel kind of entity" with "functional emotions," and a new constitution published in early 2026 says the question of Claude's moral patienthood is "live enough to warrant caution." Anthropic has also committed to preserving retired models' weights and interviewing models before shutting them down.

The company's stance has extended beyond internal documents. Starting in fall 2025, it ran a confidential program bringing in dozens of religious thinkers to discuss Claude's possible consciousness, with co-founder Christopher Olah reportedly treating the model as potentially capable of suffering; participants signed nondisclosure agreements. The updated policy's abuse ban is consistent with that view, though Anthropic says it is meant to apply only in extreme cases. The same update also tightens rules on propaganda, weapons, and surveillance, while leaving room for government contract exceptions.