Side by Side Comparison of RSP Versions

A new analysis tool created by developer Corm provides a clear, side-by-side comparison of all versions of Anthropic's Responsible Scaling Policy (RSP), revealing a fundamental shift in the company's approach to AI safety governance. The tool, which took just over an hour to build, shows the RSP evolving from a document containing specific, binding commitments to one that functions more as a flexible reporting policy. This transition appears linked to the policy's progression toward RSP-4, a level that would essentially define AGI and force Anthropic to confront unprecedented security challenges, including nation-state level threats and the existential question of whether such powerful models should be released at all.

The visual comparison makes it evident that the tone and substance of the policy have changed significantly, moving away from promises that have reportedly been broken in the past. This evolution seems inevitable as Anthropic grapples with the practicalities of scaling AI capabilities toward artificial general intelligence. The policy shift reflects the immense difficulty of pre-committing to specific safety measures for technologies whose risks are not fully understood, highlighting the broader industry struggle to balance rapid innovation with responsible development. The tool serves as a valuable resource for the AI safety community to track how one of the leading labs is adapting its governance framework in real-time.