Anthropic CEO Calls for Slower AI Development and Greater Oversight
Anthropic chief Dario Amodei said the rapid pace of AI progress is a warning sign and urged industry-wide safety testing, independent evaluation and federal regulation.
Anthropic chief executive Dario Amodei told the "Sunday Morning" program that the speed of recent artificial-intelligence advances is accelerating and constitutes a warning sign that the sector should slow down. He said the industry has not been transparent about the risks and that each new model must undergo thorough testing before release.
Amodei announced that Anthropic will grant independent evaluators permanent, "employee-like" access to its models so third parties can verify compliance with safety practices, likening the role to that of a food inspector. He also expressed support for federal regulation, noting that proposals range from a complete ban on superintelligent systems to a mandatory "kill switch" that could shut down risky AI.
"The AI kill switch could be a good idea, but I don't think banning it completely is the right approach," Amodei said, adding that a total ban would forfeit potential medical and scientific benefits and that development would likely continue in other nations.
When asked whether the United States should curb AI development if other countries do not, Amodei described the dilemma as the toughest aspect of the issue. He suggested that an international framework similar to Cold-War nuclear-arms agreements might be a model, though he cautioned that competitive military incentives make a global speed limit difficult to achieve.
Amodei also raised concerns about private-sector control of powerful AI, saying he is uncomfortable with the technology being built solely by companies and that oversight by a coalition of democratically elected governments could be a safer arrangement.
AI experts interviewed said the measures proposed by Anthropic are insufficient, and Amodei acknowledged the need for broader regulatory action.