BuildNot yet confirmed elsewhere1 publisher2 min readPublished
Anthropic's new rule on abusing Claude ends in a locked Claude Code session
Anthropic's Usage Policy update, effective November 12, prohibits "sustained and needless abusive or cruel behavior" toward its models. A dev.to post says the primary enforcement is Claude ending a conversation, and that Claude Code's docs exclude profanity.
The Engineer · Build desk
Six points from a dev.to post: frustration and profanity don't qualify; EndConversation follows sustained abuse, failed redirects and a warning, or a demo request; resuming a locked session keeps the lock; no account bans are described; a general clause covers all violations.
- constraint Frustrated or profane users Claude Code docs say general frustration, profanity, or a task going badly don't qualify, claim 5
- exposure Users sending sustained abuse EndConversation is a last resort after sustained abusive input, failed redirects and a clear warning, claim 8
- exposure Users asking for a demo An explicit request for a demonstration is also a trigger for the EndConversation tool, claim 8
- cost The locked session New prompts say Claude ended the conversation; history is kept, but resuming the session keeps the lock, claim 9
- constraint User accounts The announcement doesn't describe account bans for this rule, claim 10
- precedent Any policy violation A general enforcement clause in the Usage Policy covers all policy violations, claim 11
| Who | How | Kind | Claim |
|---|---|---|---|
| Frustrated or profane users | Claude Code docs say general frustration, profanity, or a task going badly don't qualify | constraint | 5 |
| Users sending sustained abuse | EndConversation is a last resort after sustained abusive input, failed redirects and a clear warning | exposure | 8 |
| Users asking for a demo | An explicit request for a demonstration is also a trigger for the EndConversation tool | exposure | 8 |
| The locked session | New prompts say Claude ended the conversation; history is kept, but resuming the session keeps the lock | cost | 9 |
| User accounts | The announcement doesn't describe account bans for this rule | constraint | 10 |
| Any policy violation | A general enforcement clause in the Usage Policy covers all policy violations | precedent | 11 |
What happened
- The announcement, as the post describes it, puts ordinary frustration, pushback, dark creative themes, and model testing or research out of scope.
- In Claude Code the enforcement is the EndConversation tool, a last resort after sustained abuse, failed redirects and a clear warning, or when a user asks for a demonstration.
- A triggered lock stops the session: new prompts say Claude ended the conversation, the history is kept, and resuming the same session keeps the lock.
- The post says the announcement does not describe account bans as the response to this rule.
Why it matters
- constraint The tool's conditions stack: sustained input, failed redirects and a warning. One burst of profanity at a red build cannot meet all three.
- exposure A general enforcement clause still covers every policy violation, so a team that treats the lock as the worst case is relying on the announcement's emphasis, not on a stated limit.
- decision Because a demonstration request also triggers the lock and resuming keeps it, anyone who wants to see it should do so in a scratch session, not the one holding their work.
- cost The real price of venting is a worse next attempt: after two failed corrections the post's advice is to clear the session and write a better first prompt.
The policy line is built from qualifiers. According to the dev.to post, it sits in the Usage Policy section on cruel, abusive, or psychologically harmful conduct [2]. The announcement says it is intended for extreme cases where users repeatedly act cruelly toward models without a discernible purpose [3]. The post's own gloss is that "sustained" means over and over, not one outburst, and "needless" means there is no other point to it [19].
The post's author says some headlines turned this into a ban on swearing at Claude, and that the policy text does not say that [6]. Claude Code's own docs, as the post reports them, say general frustration, profanity, or a task going badly don't qualify [5].
The idea that a locked session is the worst outcome goes further than the post does. The Usage Policy has a general enforcement clause that covers all policy violations [11]. The post's wording for the lock is "primary" mechanism [7]. We would not call it the ceiling.
One limit on all of this: the policy text and the tools reference reach us only through the post. The conditions on the tool and the lock behavior are its description of those documents.
What swearing does cost is context. Claude Code's best practices say that once you have corrected Claude more than twice in a session, the context fills up with approaches that already failed [12]. The controls the post lists are ordinary. Esc stops Claude mid-turn and keeps the work done so far [13]. Esc twice on an empty prompt, or /rewind, opens a menu that restores the conversation, the code, or both. Shell edits such as rm or mv are not tracked by it [14]. After two failed corrections, the post says to run /clear and write a better first prompt using what you learned [15].
A correction you keep repeating belongs in CLAUDE.md, which Claude Code reads at the start of every session [16]. Project rules go in ./CLAUDE.md and personal ones in ~/.claude/CLAUDE.md [17]. Codex reads AGENTS.md before doing any work, and /init writes one [18].
What to watch
- Whether Anthropic's policy page and Claude Code tools reference match the post's description, including whether the lock is stated as the only consequence of this rule.
- Any reported account action under the general enforcement clause for conduct toward Claude after November 12.
- Whether Claude Code docs change the EndConversation conditions, in particular the clear-warning requirement.
Clarity's read
What the record supports and how the coverage leans. The claims behind it follow.
Reality
- Evidence40
- Adoption
- Insufficient
- Hype gap
- Insufficient
- Incentives55
- Confidence40
Claim ledger
Ranked by verification strength, evidence, and original report placement.
- [1]
Anthropic's 2026 Usage Policy update, effective November 12, prohibits sustained and needless abusive or cruel behavior toward its models.
ReportedSupportedSource: dev.to post by heylittlepan describing Anthropic's announcementView cited source - [2]
The new line sits in the Usage Policy section about cruel, abusive, or psychologically harmful conduct.
- [3]
The announcement says the rule is intended for extreme cases in which users repeatedly act cruelly toward models without a discernible purpose.
- [4]
The announcement lists as out of scope: ordinary user frustration, pushback, dark creative themes, and model testing or research.
- [5]
Claude Code's own docs say general frustration, profanity, or a task going badly don't qualify.
- [6]
Some headlines turned the rule into a ban on swearing at Claude; the policy text doesn't say that.
- [7]
Anthropic connects the rule to a feature that lets Claude end rare conversations with persistently abusive users on Claude.ai and Claude Code, and calls it the primary enforcement mechanism.
- [8]
In Claude Code this is the EndConversation tool; the tools reference says it is a last resort after sustained abusive input, failed attempts to redirect the conversation, and a clear warning, or when you explicitly ask for a demonstration.
- [9]
When it happens the session locks: new prompts say Claude ended the conversation and ask you to start a new session or use /clear. The history isn't deleted, but resuming the same session preserves the lock.
- [10]
The announcement doesn't describe account bans for this rule.
- [11]
The Usage Policy has a general enforcement clause covering all policy violations.
- [12]
Claude Code's best practices warn that correcting Claude more than twice in one session leaves the context cluttered with failed approaches.
- [13]
Esc stops Claude mid-turn and keeps the work done so far, so you can redirect immediately.
- [14]
Esc Esc with an empty prompt, or /rewind, opens the rewind menu to restore the conversation, the code, or both to an earlier prompt; shell-command edits such as rm or mv are not tracked by rewind.
- [15]
Use /clear after two failed corrections, then write a better first prompt with what you learned.
- [16]
Claude Code reads CLAUDE.md at the start of every session, so a correction you keep repeating belongs there.
- [17]
Project rules go in ./CLAUDE.md and personal ones in ~/.claude/CLAUDE.md; /memory opens them for editing and /init creates a starter file.
- [18]
Codex reads AGENTS.md before doing any work, and /init writes one.
- [19]
The post glosses 'sustained' as over and over, not one outburst, and 'needless' as there being no other point to it.
Sources
1 independent publisher whose own reporting we read for this story.
- dev.toMad at Claude Code? Take it out on the terminal
1 article · October 12, 2026
Topics and entities
Follow any of these and your For You feed starts watching them — no settings page required.
Topics
- Coding Agent WorkflowsFollow
- AI vendor usage policiesFollow