CONCEPT
Responsible Scaling Policy
Anthropic's framework of capability thresholds — AI Safety Levels analogous to biosafety levels — specifying safety measures required before deployment at each level, designed to build the governance framework before the harm rather than after.
The Responsible Scaling Policy (RSP) is Amodei's attempt to build a prospective governance framework for AI deployment — establishing the institutional structures for managing risk at each level of capability before the capability is achieved. Structured around a series of capability thresholds called AI Safety Levels, analogous to biosafety levels in pathogen research, the framework specifies safety measures that must be in place before a system can be deployed at scale at each level. The RSP embodies three principles that distinguish it from typical technology governance: capability and safety are evaluated together, evaluation is prospective rather than reactive, and the framework is binding on the organization rather than advisory. The framework addresses risks including biological and chemical weapons assistance, cyberattack capability, autonomous behavior, and harmful outputs.
In The You On AI Field Guide
The history of powerful technologies is a history of missing frameworks. Nuclear energy arrived before its regulatory infrastructure. The automobile arrived before traffic