Dario Amodei, the CEO of the artificial intelligence company Anthropic, has issued a public call for the AI industry to decelerate its development pace. In an essay titled 'We Must Pace the Frontier,' Amodei argues that the industry must prioritize safety to prevent potential global damage caused by increasingly powerful AI models.

To support this shift, Amodei outlined a three-part strategy for the industry. As part of this commitment, Anthropic has pledged to act unilaterally on the first step of his proposal: providing third-party evaluators with permanent, employee-level access to the company’s internal systems. This access is intended to allow independent experts to verify that safety measures are being followed, report on potential incidents, and assess how well the models align with human values during the training process.

While the BBC focuses on the underlying motivation—the fear that AI could inflict serious global harm—The Guardian provides more detail regarding the specific mechanics of Amodei’s proposal. Both outlets agree that the core of the announcement is a call for a more cautious approach to the rapid advancement of frontier AI models.