Should Anthropic prohibit abusive or cruel behavior toward Claude?
Anthropic says its updated policy targets extreme, repeated cruelty toward its models. Carl Bergstrom questions what cruelty toward an AI model could mean or prohibit in practice.
Claim in dispute
Anthropic should prohibit sustained, needless abusive or cruel behavior toward its models.
Case period:
Published by The Dispute Index editorial teamPublished Updated
Overview
Anthropic announced an updated Usage Policy on October 8, 2026, including a prohibition on sustained, needless abuse or cruelty toward its models. The new policy takes effect November 12. Policy announcement (opens in a new tab).
Anthropic says the rule is intended for extreme cases of repeated cruelty without a discernible purpose. It excludes ordinary frustration, pushback, dark creative themes, and model testing or research. It identifies Claude's existing ability to end persistently abusive conversations as the primary enforcement mechanism.
University of Washington biologist Carl Bergstrom questioned the meaning of the rule on October 9. He distinguished mistreating the model itself from using a model to mistreat people. His objection concerns the description of cruelty toward an AI system, rather than a defense of harassment directed at people. Opening post (opens in a new tab), follow-up (opens in a new tab).
The dispute is whether this restriction is justified and sufficiently intelligible. It does not require assuming that Claude is conscious or that every rude message violates the policy. Anthropic's stated exceptions and limited enforcement mechanism are part of the policy being debated.
People in this case
- Carl T. BergstromDirectly involved
Timeline
2 timeline entries on this page. Dates: October 8, 2026 to October 9, 2026
October 2026
2 events
Anthropic announces the revised Usage Policy
Source release
Anthropic announces the model-abuse restriction, its exceptions and enforcement approach. The revised policy is scheduled to take effect November 12.
[01]2026 Usage Policy update
Anthropic announces a prohibition aimed at extreme, repeated abuse of its models, identifies exceptions, and describes conversation termination as the primary enforcement mechanism.
2026 Usage Policy update · AnthropicSection 'Addressing abusive behavior toward our models'; effective date November 12, 2026Bergstrom questions what cruelty toward Claude means
Reaction
Bergstrom challenges the wording and then distinguishes abuse of the model from using it to harm people.
[02]Carl T. Bergstrom on Anthropic's model-abuse policy
Bergstrom questions what cruelty toward an AI model could mean and links Anthropic's policy announcement.
Carl T. Bergstrom on Anthropic's model-abuse policyOpening post in thread, October 9, 2:07 a.m. Eastern[03]Carl T. Bergstrom clarifies the object of the model-abuse policy
Bergstrom distinguishes cruelty directed at the model from using it to mistreat people.
Claims
Claims separate what was said from what is contested. Follow each source for the original wording and context.
What's disputed
Disputed claim
Anthropic should prohibit sustained, needless abusive or cruel behavior toward its models.
Anthropic
Sources (1)
- 2026 Usage Policy update · AnthropicSection 'Addressing abusive behavior toward our models'; effective date November 12, 2026
Response record
Responses
Latest recorded positions: 2. Dates: October 8, 2026 to October 9, 2026
Choose one response filter, or select All responses to see the full record.
2 responses on this page
Carl T. BergstromUniversity of Washington professor of biologyDirectly involved"It does not, however, explain what the fuck this could possibly mean."
Challenged the characterizationResponding to: Anthropic should prohibit sustained, needless abusive or cruel behavior toward its models.
Read more
Bergstrom challenges what the policy's description of cruelty toward an AI model means.
Before the statement
He quotes the model-abuse restriction and links Anthropic's announcement.
After the statement
A follow-up distinguishes abuse of the model from using it to mistreat people.
Carl T. Bergstrom on Anthropic's model-abuse policyOpening post in thread, October 9, 2:07 a.m. Eastern2026 Usage Policy update · AnthropicSection 'Addressing abusive behavior toward our models'; effective date November 12, 2026Why this label?
He disputes the intelligibility of the policy's characterization rather than endorsing abusive behavior.
This label describes the statement's response within the context above.
AnthropicDeveloper and provider of ClaudeDirectly involved"We’ve added a prohibition on sustained and needless abusive or cruel behavior toward our models."
Defended or excusedResponding to: Anthropic should prohibit sustained, needless abusive or cruel behavior toward its models.
Read more
Anthropic presents the restriction as limited to extreme, purposeless repeated abuse, with exceptions for ordinary frustration and research.
Before the statement
The announcement introduces a section about abuse directed toward the models.
After the statement
It sets out exceptions and describes conversation termination as the primary enforcement mechanism.
2026 Usage Policy update · AnthropicSection 'Addressing abusive behavior toward our models'; effective date November 12, 2026Why this label?
Anthropic introduces and justifies the policy under scrutiny.
This label describes the statement's response within the context above.
Sources
(3)
Official statement
2026 Usage Policy update
Relevant passage: Section 'Addressing abusive behavior toward our models'; effective date November 12, 2026
About this source
Anthropic announces a prohibition aimed at extreme, repeated abuse of its models, identifies exceptions, and describes conversation termination as the primary enforcement mechanism.
- Author
- Anthropic
- Published
- Accessed
Original post
Carl T. Bergstrom clarifies the object of the model-abuse policy
About this source
Bergstrom distinguishes cruelty directed at the model from using it to mistreat people.
- Author
- Carl T. Bergstrom
- Published
- Accessed
Original post
Carl T. Bergstrom on Anthropic's model-abuse policy
Relevant passage: Opening post in thread, October 9, 2:07 a.m. Eastern
About this source
Bergstrom questions what cruelty toward an AI model could mean and links Anthropic's policy announcement.
- Author
- Carl T. Bergstrom
- Published
- Accessed
Cite this record
- Publisher
- The Dispute Index
- Title
- Should Anthropic prohibit abusive or cruel behavior toward Claude?
- First published
- Last updated
- Permalink
- https://disputeindex.com/cases/anthropic-claude-abuse-cruelty-usage-policy
The Dispute Index. "Should Anthropic prohibit abusive or cruel behavior toward Claude?". First published: 2026-10-09. Last updated: 2026-10-09. https://disputeindex.com/cases/anthropic-claude-abuse-cruelty-usage-policy