Usage Policy Update

Anthropic is updating its Usage Policy in line with the expanding capabilities of its products and advances in technology. These changes take effect from September 15, 2025.

To address agentic capabilities and cybersecurity, the policy manages the risks of agentic tools such as Claude Code and Computer Use. While creating malicious software or breaching networks remains prohibited, use for security-strengthening purposes—such as finding vulnerabilities with the system owner's consent—continues to be supported.

Restrictions on political content move away from the previous blanket ban toward a more specific approach. While strictly prohibiting actions that disrupt electoral processes or deceive voters, the scope has been adjusted to allow legitimate political discourse and research activities, such as policy research and civic education.

Standards for law enforcement and high-risk cases have also been clarified. Surveillance and profiling are prohibited, but use as back-office and analytical tools is permitted. In addition, the policy specifies that the human-in-the-loop and AI-disclosure obligations applied to high-risk cases such as legal and financial fields apply only to consumer-facing model outputs, not to business-to-business (B2B) interactions.