Anthropic has revised its usage policy for the Claude AI chatbot to address increasing AI safety concerns. The updated policy introduces stricter cybersecurity rules and explicitly forbids the use of Claude in developing high-yield explosives and CBRN (biological, nuclear, chemical, and radiological) weapons. With the launch of Claude Opus 4, Anthropic implemented “AI Safety Level 3” protections to prevent jailbreaking and assistance in creating dangerous weapons. The company also added new restrictions against using Claude for hacking, malware creation, or denial-of-service attacks, reflecting fears around misuse of agentic AI tools like Computer Use and Claude Code. While loosening rules on political content, Anthropic now prohibits uses that disrupt democratic processes or involve deceptive campaigning. The policy’s “high-risk” use requirements apply only to consumer-facing scenarios, not business uses.
Back
