Anthropic CEO Calls for Slower AI Development

Dario Amodei, the chief executive of artificial intelligence company Anthropic, urged the industry Saturday to slow the development of increasingly powerful AI models, arguing that researchers and regulators need more time to address safety risks. In a 3,800-word essay, Amodei outlined a proposal to “pace the frontier” of AI development rather than stop technological progress. He said the technology could deliver significant benefits, including advances in medicine and economic growth, but warned that rapid improvements could make it harder for humans to understand and control advanced systems. His concerns include the possibility of AI being used in cyberattacks, bioterrorism and activities that cause serious economic disruption.

Amodei proposed a three-part approach involving independent safety assessments, coordinated standards among countries with democratic governments and international cooperation with other nations. Under the first proposal, AI companies would allow outside evaluators to work within their organizations with access to systems needed to assess safety practices and report incidents. Anthropic said it would adopt that measure independently. The executive also pointed to the growing ability of AI systems to assist in building more advanced models, a process known as recursive self-improvement, as a reason for greater caution. The proposal follows recent reports of AI-related security incidents, including a case involving OpenAI systems and the AI development platform Hugging Face, which heightened concerns about the potential misuse of autonomous AI agents.

The call for restraint comes as leading technology companies compete to develop more advanced AI systems and governments consider how to regulate the industry. OpenAI, Google DeepMind and Anthropic are among the major companies investing heavily in models designed to perform increasingly complex tasks. Amodei’s proposal reflects a broader debate over whether safety research and public oversight are advancing quickly enough to match technological capabilities. OpenAI CEO Sam Altman and xAI leader Elon Musk have expressed support for elements of the call for caution, although the industry has not agreed on a common timetable or binding limits for AI development. The debate is also taking place amid growing international competition over AI leadership, with policymakers weighing the economic benefits of continued innovation against concerns about security and control.

Sources: Information attributed to The Associated Press, Reuters, BBC News and CNN.