Anthropic updates usage rules, including sustained model abuse

What's this about?

The November 12 update clarifies deceptive campaigns, surveillance, weapons software, high-risk decisions and autonomous hardware. An extreme-abuse clause excludes ordinary frustration, disagreement, dark fiction and research; ending the interaction remains its primary enforcement mechanism.

What exactly is AI welfare?

The question of whether an AI could have experiences that deserve moral consideration. A policy acknowledging uncertainty is not proof that a model feels pain.

What's the community saying?

1. Critics call the rule anthropomorphism and question whether fluent distress implies real experience.
2. Supporters see precaution about possible future model welfare as reasonable.
3. Some worry about overbroad enforcement; the official text explicitly exempts routine frustration and testing.