Anthropic has announced updates to its Usage Policy for Claude, reflecting the growing capabilities and evolving usage of its products. These changes, which will take effect on September 15, 2025, aim to provide greater clarity and detail based on user feedback, product changes, regulatory developments, and enforcement priorities.
Key Points
- The updated Usage Policy takes effect on September 15, 2025.
- A new section prohibits malicious computer, network, and infrastructure compromise activities, addressing risks from advanced agentic capabilities.
- The policy now permits legitimate political discourse and research, tailoring restrictions to specifically prohibit deceptive or disruptive activities to democratic processes or voter/campaign targeting.
- Language regarding law enforcement use has been clarified, maintaining existing restrictions on surveillance, tracking, profiling, and biometric monitoring.
- High-Risk Use Case Requirements, which mandate safeguards like human-in-the-loop oversight and AI disclosure, now apply specifically to consumer-facing model outputs, not business-to-business interactions.
- The updates follow the release of agentic tools such as Claude Code and Computer Use.
- Anthropic's first threat intelligence report, "Detecting and Countering Malicious Uses of Claude: March 2025," informed the changes related to agentic risks.
Context
According to Anthropic, the updates are a response to rapid advances in agentic capabilities, including the release of tools like Claude Code and Computer Use. These capabilities introduce new risks such as scaled abuse and cyber attacks, as detailed in their March 2025 threat intelligence report. The company also noted user feedback regarding broad prohibitions on political content and ambiguities in law enforcement use cases.
Why It Matters
These policy updates clarify the permissible and prohibited uses of Claude, particularly for developers and organizations leveraging its agentic capabilities or deploying it in sensitive domains. Builders must understand these revised guidelines to ensure compliance and mitigate risks associated with advanced AI applications.
What To Do
- Review the full updated Usage Policy on Anthropic's website to understand all changes.
- Note the effective date of September 15, 2025, for these policy changes.
- Consult the new Help Center article on agentic use for concrete examples of prohibited activities.
- Assess existing or planned Claude deployments, especially those involving agentic functions, political content, law enforcement, or consumer-facing applications, against the new guidelines.