The company has updated its usage guidelines to ban ongoing abusive interactions with its Claude model and to restrict AI-driven political campaigns that could mislead voters.
Anthropic has overhauled its usage policy for Claude, its flagship AI assistant, explicitly banning users from engaging in persistent abusive interactions and tightening rules around deceptive political applications.
Key Changes to the Abuse Policy
The revised guidelines now define "persistent abuse" as repeated attempts to provoke harmful or harassing content from Claude. Any user who tries to elicit hate speech, threats, or targeted harassment over multiple sessions will face immediate suspension of access.
Restrictions on Political Manipulation
Anthropic also narrows the permissible scope of political use. The policy prohibits employing Claude to generate misleading campaign material, fabricate voter outreach messages, or otherwise influence elections through deceptive AI‑generated content.
The company emphasizes that legitimate political analysis, fact‑checking, and policy discussion remain allowed, provided the output is clearly labeled as AI‑generated and does not aim to deceive voters.
Enforcement and Compliance
Violations will be tracked through automated monitoring and human review. Anthropic pledges to cooperate with regulators and to publish periodic transparency reports detailing enforcement actions.
- Immediate account suspension for repeated abuse attempts
- Mandatory labeling of political content generated by Claude
- Cooperation with election oversight bodies
Our goal is to keep Claude safe and trustworthy while respecting legitimate political discourse.
The updated policy reflects growing industry pressure to curb AI misuse in both harassment and election contexts, aligning Anthropic with similar moves by other AI developers.