11
NEW YORK: Anthropic CEO Dario Amodei has called for the artificial intelligence industry to slow the pace of development, warning that AI capabilities could advance to potentially dangerous levels before adequate safety measures are in place.
In a post published Saturday, Amodei warned that within six to 12 months, AI systems could become capable of carrying out highly sophisticated cyber operations, including potentially coordinating large-scale attacks. He argued that even a temporary slowdown could provide critical time to strengthen safeguards and reduce the risks associated with increasingly powerful AI systems.
Amodei proposed a series of measures aimed at improving oversight of frontier AI development. Among them is a recommendation that leading AI companies provide independent external evaluators with continuous access to their operations so that safety practices can be monitored more effectively.
He said Anthropic is prepared to implement such a system by providing external evaluators with office access, company laptops and other resources needed to assess its safety practices.
The Anthropic chief also called for greater cooperation between AI companies and governments. He suggested that the U.S. government consider legal waivers allowing American AI companies to coordinate on safety standards without violating antitrust regulations.
Amodei further urged democratic governments to coordinate internationally, including with authoritarian states, to prevent companies in different countries from accelerating AI development independently while others attempt to introduce safety-related pauses.
His warning comes amid growing concerns about the misuse of increasingly capable AI systems. Anthropic said this week that it had blocked attempts to exploit its models for cyberattacks, surveillance, propaganda and research that could potentially contribute to biological weapons development.
The debate has also intensified following the resignation of Anthropic researcher Jacob Coxon, who warned that competition among leading AI companies could be placing technological advancement ahead of adequate safety precautions.
Amodei acknowledged that the measures he proposed would be difficult to implement but argued that slowing the development of frontier AI systems, even temporarily, could provide valuable time to advance safety research and reduce the possibility of catastrophic outcomes.
