Dario Amodei, chief executive of AI developer Anthropic, issued an appeal on Saturday urging the artificial intelligence sector to “slow down” its rapid pace of innovation. He accompanied this call to action with a three-part strategic plan aimed at enhancing safety and oversight within the industry.
Amodei published the details of his proposal in an essay titled “We Must Pace the Frontier,” which was linked via a social media post. The plan outlines specific measures to ensure that AI systems remain aligned with safety standards as they become more powerful.
Highlighting his commitment, Amodei stated that Anthropic would “unilaterally” commit to the first step of the proposed plan without waiting for industry-wide consensus. This initial measure involves granting third-party evaluators permanent, employee-level access to Anthropic’s systems. Through this access, independent auditors would be able to verify adherence to safety protocols, report on any incidents, and assess model alignment during the training process.
Third-party access sounds good in theory, but how do we ensure auditors are truly independent?
I’m surprised Amodei is asking for a slowdown. Didn’t we want faster progress? What changed?
Unilateral action is bold, but will competitors actually care? Seems like a PR move without teeth.
Finally, a CEO willing to put safety before speed. Hope other labs follow Anthropic’s lead here.