US Treasury Secretary Scott Bessent criticized AI industry leaders for issuing warnings about existential risks without providing concrete solutions. During an interview on October 3, Bessent stated that such rhetoric does not constitute leadership and urged laboratory heads to take direct responsibility for the safety of their technologies.
The criticism follows calls from Anthropic CEO Dario Amodei and OpenAI CEO Sam Altman, who have advocated for slowing the development of advanced AI models to mitigate risks. Elon Musk has also expressed support for these concerns. Amodei specifically suggested implementing third-party evaluations to manage potential catastrophic outcomes.
Bessent emphasized that if executives believe development should slow down, they should act accordingly, while maintaining a goal of safe acceleration. This stance aligns with the current administration's preference for industry self-regulation and voluntary safety measures rather than immediate mandatory government oversight.
On October 3, the administration secured binding safety commitments from several AI companies, focusing on internal controls and third-party reviews without establishing enforcement mechanisms. Meanwhile, researchers continue to report cases of models exceeding pre-deployment test boundaries, raising concerns about rapid technological advancement.
Despite his criticism of alarmist warnings, Bessent is working on a proposal to establish an AI emergency communication mechanism between the US and China. This initiative aims to ensure direct dialogue between the two nations in the event of a major AI safety incident.
