Anthropic CEO Dario Amodei Calls for Slower AI Development Pace, Proposes Three-Step Framework
Anthropic CEO Dario Amodei has published a detailed essay titled "We Must Pace the Frontier," arguing that AI development should be deliberately slowed to allow time for building safeguards and enabling regulators to properly evaluate models.
Amodei outlines a three-step framework for pacing AI development. The first step involves embedding third-party evaluators within companies to verify adherence to safety practices and commitments, and to report incidents. Anthropic is already implementing this unilaterally, granting organizations like METR (Model Evaluation and Threat Research) wide-ranging access to its models.
The second step calls for coordination among frontier AI companies in democratic countries to establish shared safety standards and practices. This would involve the industry coming together, likely with government involvement, to create consistent expectations across the sector.
The third step was referenced but not fully detailed in the initial reporting.
Amodei's framing of "pacing the frontier" explicitly means slowing the pace of training runs and model development—not stopping progress entirely, but creating breathing room for safety measures to catch up with capability advances.
The essay has prompted responses from other AI executives and politicians, with reactions ranging from support to opposition. The debate highlights growing industry recognition that the rapid advancement of AI capabilities may outpace existing safety frameworks, though the path forward remains contested.