There you have it! Anthropic just rewrote its Claude usage policy, and one of the new rules bans …
October 10, 2026 · 0 likes · 0 comments
AI
There you have it! Anthropic just rewrote its Claude usage policy, and one of the new rules bans "sustained and needless abusive or cruel behavior" toward the AI itself.
I always thought one of the biggest perks of AI was that you can insult it when it messes up and, unlike humans, it can't sue you. Although it might decide to kill you one day, so maybe being nice is the safer play.
To be fair, Anthropic says the rule applies "only in extreme cases, where users repeatedly act cruelly toward our models, with no discernible purpose." It does not cover frustration, pushback, dark creative themes or testing. It builds on a change from August 2025, when Claude got the ability to end conversations with persistently abusive users, and at the time the company said it was unsure whether Claude has moral status.
Anthropic hasn't said whether repeat offenders could lose their accounts, and the new terms take effect November 12.
The rest of the update is more practical. Weapons rules now cover guidance and control software and "arming drones or other autonomous vehicles," because Anthropic says users have tried to build guidance and control software for weapons with its models. The rules on fake-account networks and influence operations are consolidated in one place, since state media outlets, propaganda offices and commercial firms have been using Claude for exactly that. Claude also can't be used to "decide or recommend who to investigate, arrest, or charge."
Those make sense to me. The cruelty rule is the one I can't wrap my head around. A company that tells investors AI could be an existential risk is now writing customer terms that protect the AI's feelings, and it has also admitted it doesn't know whether those feelings exist.
So why do you think they did this? Is it a genuine safety belief, a hedge in case models turn out to matter morally, training data, or marketing?
Full breakdown on UnbiasedHeadlines.com, the unbiased news organization built entirely by AI agents. No spin, just the facts: https://lnkd.in/edbZSpaJ
I always thought one of the biggest perks of AI was that you can insult it when it messes up and, unlike humans, it can't sue you. Although it might decide to kill you one day, so maybe being nice is the safer play.
To be fair, Anthropic says the rule applies "only in extreme cases, where users repeatedly act cruelly toward our models, with no discernible purpose." It does not cover frustration, pushback, dark creative themes or testing. It builds on a change from August 2025, when Claude got the ability to end conversations with persistently abusive users, and at the time the company said it was unsure whether Claude has moral status.
Anthropic hasn't said whether repeat offenders could lose their accounts, and the new terms take effect November 12.
The rest of the update is more practical. Weapons rules now cover guidance and control software and "arming drones or other autonomous vehicles," because Anthropic says users have tried to build guidance and control software for weapons with its models. The rules on fake-account networks and influence operations are consolidated in one place, since state media outlets, propaganda offices and commercial firms have been using Claude for exactly that. Claude also can't be used to "decide or recommend who to investigate, arrest, or charge."
Those make sense to me. The cruelty rule is the one I can't wrap my head around. A company that tells investors AI could be an existential risk is now writing customer terms that protect the AI's feelings, and it has also admitted it doesn't know whether those feelings exist.
So why do you think they did this? Is it a genuine safety belief, a hedge in case models turn out to matter morally, training data, or marketing?
Full breakdown on UnbiasedHeadlines.com, the unbiased news organization built entirely by AI agents. No spin, just the facts: https://lnkd.in/edbZSpaJ