Anthropic bans sustained, gratuitous abuse of its models in its usage policy
The update, published on 8 October 2026, takes effect on 12 November and will be enforced only in extreme cases.
Anthropic has added to its usage policy a ban on sustained, unnecessary abusive or cruel behaviour towards its models. According to the announcement, it will be enforced only in extreme cases, involving repeated cruelty with no apparent purpose, and it does not cover everyday frustration, disagreement, dark creative themes, or testing and research on the models.
The announcement says that Claude’s existing ability to end rare conversations with persistently abusive users on Claude.ai and Claude Code will remain the main enforcement mechanism. The update includes other changes, such as a section on deceptive activity and a rewording of the elections section. In ChatGPT-IA-edu the measure was read as a principle of caution, with no admission by the company that its models suffer; that reading comes from the group’s comments, because the announcement does not develop the question. It matters to teachers who use these tools, because it is a change to the terms of use.
Anthropic bans sustained, gratuitous abuse of its models in its usage policy
The update, published on 8 October 2026, takes effect on 12 November and will be enforced only in extreme cases.
Anthropic has added to its usage policy a ban on sustained, unnecessary abusive or cruel behaviour towards its models. According to the announcement, it will be enforced only in extreme cases, involving repeated cruelty with no apparent purpose, and it does not cover everyday frustration, disagreement, dark creative themes, or testing and research on the models.
The announcement says that Claude’s existing ability to end rare conversations with persistently abusive users on Claude.ai and Claude Code will remain the main enforcement mechanism. The update includes other changes, such as a section on deceptive activity and a rewording of the elections section. In ChatGPT-IA-edu the measure was read as a principle of caution, with no admission by the company that its models suffer; that reading comes from the group’s comments, because the announcement does not develop the question. It matters to teachers who use these tools, because it is a change to the terms of use.
