Anthropic, an artificial intelligence research company, has updated its Responsible Scaling Policy (RSP) to address the risks posed by increasingly powerful AI models. The RSP lays out a framework for ensuring that AI models are developed and deployed responsibly, with safeguards in place to prevent catastrophic harm. The company's approach is based on a tiered system of safety levels (ASLs), with higher levels requiring more stringent safeguards for models with greater capabilities. The RSP also introduces the concept of Capability Thresholds, specific levels of AI capability that trigger the implementation of increased safeguards. The policy includes detailed procedures for assessing model capabilities, implementing appropriate safeguards, and ensuring transparency and accountability in the development and deployment of AI systems.