Anthropic's usage policy update: what changes for Claude users on November 12, 2026

The Morse teamPublished November 7, 20269 min read

The Anthropic usage policy update takes effect on November 12, 2026, replacing the version in force since September 15, 2025. Most of it rewrites rules for developers and businesses, but two changes reach anyone who chats with Claude: a new ban on "sustained and needless abusive or cruel behavior toward our models," and an enforcement line that now starts "If we suspect that you may have violated our Usage Policy" where the old one said "If we learn that you have violated" it.

On November 7, 2026, we read the September 2025 text in Anthropic's archive and the November 12 text side by side, along with Anthropic's October 8 announcement, its help pages on warnings, appeals, and retention, its account of how its safeguards work, and its privacy policy. Anthropic deserves credit for how it handled this one. It posted the new text five weeks early, kept the old one online, and its announcement says which rewrites don't change what it enforces in practice.

What changed in Anthropic's usage policy?

The policy grew from 14 universal standards to 15, and its high-risk section was rewritten. As of November 7, 2026, the new text is posted, but the September 2025 version governs until November 12.

Two smaller edits ease the rules for businesses. Both versions require them to tell people when AI shaped advice or a decision about them, and the new one adds that "users do not need to identify Anthropic, Claude, or the model used." The new high-risk section also lets a business skip human review for decisions "wholly favorable to the individual," like paying an insurance claim.

Can Anthropic ban you for being rude to Claude?

Not for being rude. Anthropic's announcement says the rule is "meant to apply only in extreme cases, where users repeatedly act cruelly toward our models, with no discernible purpose," and "does not apply to common versions of user frustration, pushback, dark creative themes, or model testing and research."

On October 8, Dexerto's headline said "Claude users can be banned" for it, though its story noted the company "didn't say whether repeat offenders could lose their accounts" (Dexerto). The announcement itself says "Claude's ability to end these interactions will remain the primary enforcement mechanism." Anthropic announced that ability on August 15, 2025, when it let Claude end "a rare subset of conversations." When it does, you can't send new messages in that chat, but your other chats are unaffected, you can start a new one at once, and you can "edit and retry previous messages." Claude is told not to end a chat where a user "might be at imminent risk of harming themselves or others."

Still, the rule sits among the 15 universal standards, so the policy's general line covers it like any other: on a suspected violation, Anthropic "may warn you or throttle, limit, suspend, or terminate your access" (Anthropic). The first reason Anthropic's appeals page gives for banning an account is "Repeated violations of our Usage Policy." Read together, an ended chat is the expected outcome and an account action is possible but not described for this rule.

How does Anthropic detect usage policy violations?

With Claude itself. Anthropic's account of its safeguards, published August 12, 2025, says detection runs on "prompted or specially fine-tuned Claude models called 'classifiers,'" several at a time, "each monitoring for specific types of harm while the main conversation flows naturally." On Anthropic's own apps, uploaded images are also compared against hashes of known child sexual abuse material. A classifier can quietly add instructions to Claude's system prompt to steer a reply, or in "a narrow set of cases" stop Claude from answering at all.

The same post describes "hierarchical summarization," which condenses individual chats into summaries and analyzes them "to identify account-level concerns." It says Anthropic uses "a combination of automated systems and human review." Its help article on user safety, updated March 16, 2026, adds "enhanced safety filters" that Anthropic "may temporarily apply" to users who repeatedly violate its policies, and admits the systems "are not failsafe," with "false positives or false negatives."

What happens to a Claude chat that gets flagged?

It's kept longer, can be read by designated staff, and can train Anthropic's safety systems even with training off. As of November 7, 2026, Anthropic's retention page says Anthropic keeps the inputs and outputs of a chat "flagged by our automated trust and safety systems as violating our Usage Policy" for up to two years, and safety classification scores for up to seven. The privacy policy, effective September 10, 2026, says Anthropic removes your user ID from flagged chats to train its classifiers, "however, we may re-identify the Inputs or Outputs to enforce our Terms of Service or Usage Policy."

None of this is new on November 12. The retention rule points to the Usage Policy, so from that date a chat flagged under any of the 15 new standards falls under it. Anthropic hasn't said whether its classifiers look for abuse aimed at Claude. A blocked reply on its own is not a finding against you, the new text says. How these periods compare with other Claude data is in is Claude private, and how other companies' reviewers work is in who reviews AI chats.

What happens if your Claude account is banned?

You appeal from inside the banned account. Anthropic's appeals page, updated October 30, 2026, says to "log in with your banned account, and fill out the appeal form," adding "You must be logged in to access the appeal form." A banned user on a personal plan can still export their data or delete the account, but "the data you're able to access and export may be restricted depending upon the Usage Policy violation." If you think a warning was a mistake, the page says to email Anthropic with your account details.

The page lists three grounds for a ban: repeated Usage Policy violations, signing up "from an unsupported location," and breaking the terms of service. The new policy forbids getting around a ban with another account, including "providing access to a person or entity that was previously banned," as the old one did. It adds "circumvention of the account sign-up or log-in process" to its list of bypasses.

What does the new policy say about tracking people?

It bans far more of it, in plainer words. The old surveillance rule fit in one line about location, emotional state, and communication. The new one bans targeting or tracking a person or their property without consent "by monitoring their physical location, biometrics, movements, online activities, communications, or associations, whether in real time, on an ongoing basis, or through analysis of previously collected data" (Anthropic). It also bans unmasking or locating people "from anonymous or pseudonymous activity" to harass or harm them, and building tools "designed for the surveillance uses prohibited in this section."

It allows tracking people agreed to, "such as for fraud detection or anti-money laundering screening," along with content moderation, journalism, and legal research by police or courts. Anthropic's announcement says the rewrite "does not reflect a change in what we enforce in practice." The rule protects you from someone else pointing Claude at your life. It governs users, not Anthropic, so it doesn't change the classifiers and review described above.

Does the usage policy apply when you use Claude through another app?

Yes. The November 12 text applies to "the end users of products or services integrating Claude," along with people who reach Claude through cloud providers and authorized resellers. A message you send to Claude from any app is held to the same 15 standards. What changes from app to app is whether Anthropic knows who sent it. We compared six routes in how to use Claude anonymously.

How Morse keeps your Claude chats from following you

A flagged Claude chat can sit on Anthropic's servers for up to two years, and Anthropic can re-identify it to the account that sent it. Morse is built so the message reaching Anthropic doesn't come from an account of yours. Claude runs on Morse in Anonymous mode, and Anthropic's Usage Policy still applies to every word you send.

• Anonymous, for Claude without your name on what its classifiers read. Morse removes your account, email, and IP address, then sends the message under Morse's own account. Anthropic sees what you wrote without knowing it's you, and may keep it for a limited time to check for abuse, under its own policy. Leave names and details you'd rather keep to yourself out of the message, because the words still reach Anthropic. Claude Haiku 5.5 is on the Free plan, and Claude Sonnet 5.5 comes with Pro.
• Private, for questions you'd rather no model host kept. It's the default for a new chat, with open-weight models like DeepSeek, Kimi, and GLM, plus xAI's Grok. Morse has zero data retention agreements with its inference, search, and other providers, so they keep nothing. Where a provider confirms zero retention on each reply, Morse checks it and refuses a reply that doesn't.
• E2EE, for anything no company's classifiers should read. Your browser checks the secure enclave's hardware attestation, then encrypts the message so only that enclave can open it, and Morse's servers relay ciphertext they can't read. It runs open-weight models, rests on the enclave hardware and attestation working as designed, and comes with Pro, at $19 a month.

In every mode, your chat history is stored in your browser, encrypted, and Morse's servers don't keep your messages or log what you write. Morse doesn't train AI on your chats, sell your data, or show ads. Because we don't store chats, we usually have very little to give authorities. Clearing your browser or deleting your account deletes your chats for good, and Morse can't recover them.

Send a first message from the Morse home page without an account, or start free with no card and open a Claude chat in Anonymous mode.