Last summer, Anthropic published their Responsible Scaling Policy (RSP) focusing on addressing safety failures and misuse of frontier models. The goal is to turn high-level safety concepts into practical guidelines for tech organizations. The policy identifies catastrophic risks and clarifies organizational priorities. They aim to move towards established best practices and regulations, drawing insights from various safety domains. Anthropic is committed to identifying, testing, and responding to Red Line Capabilities in models to ensure safety and security. They are actively exploring ways to integrate practices from different risk management areas. The policy also focuses on threat modeling, evaluations, and implementing the ASL-3 Standard for model safety.
https://www.anthropic.com/news/reflections-on-our-responsible-scaling-policy