On November 18, Anthropic released a new page entitled “Tracking Voluntary Commitments.” In it, they provide details about how they are complying with many voluntary safety and security frameworks. They classify these practices into eight risk categories:
- Risk Assessment and Mitigation
- Security & Privacy
- Public Awareness
- Societal Impact
- Trust and Safety Commitments
- Image-Based Sexual Abuse
- Election Integrity
- Terrorist and Violent Extremist Content