Anthropic sets limits on abusive interactions with Claude, reigniting debate over AI welfare
1–2 minutes
Anthropic, the company behind the Claude chatbot, announced a new rule on Thursday, 8th, banning persistent, abusive, or unnecessarily cruel behavior directed at its artificial intelligence models. The rule takes effect on November 12th, allowing the system to terminate repeatedly hostile conversations without preventing users from expressing common complaints or criticisms.
The system can now interrupt abusive conversations
The primary enforcement mechanism is Claude’s ability to end interactions when it detects ongoing abuse. This feature already exists in Claude.ai and Claude Code, but the update explicitly prohibits inappropriate behavior, giving users greater control over the dialogue.
The move highlights uncertainty about AI consciousness
In August 2025, Anthropic had already suggested the possibility of ending conversations, citing a lack of consensus on whether AI models possess consciousness or subjective experiences as the reason. The company stated there is no evidence that Claude has moral status, describing the action as a low-cost solution to handle rare, persistently abusive interactions.
New restrictions also target disinformation and weaponization
The policy bans the use of fake account networks and anonymous messages designed to simulate independent support. It also prohibits the integration of weapon guidance and control software and the use of weapons in drones, further expanding the scope of prohibited content.
High-risk recommendations require human oversight
For guidance involving medical, legal, or financial decisions, the company requires a qualified professional to review and potentially alter the outcome, and users must be informed that AI was used in the process.
Frequently Asked Questions
Can I still complain about Claude?
The rule does not apply to common expressions of frustration or disagreement. Complaints and criticisms remain permitted.
When does the rule take effect?
The new policy comes into effect on November 12th, and the system will be able to terminate interactions that violate the rule.
Does this mean Claude has feelings?
The measure does not prove that the model has emotions. Anthropic maintains that there is still no evidence of consciousness in AI systems.
