Anthropic CEO urges AI development slowdown to mitigate human risks

Dario Amodei, CEO of the artificial intelligence firm Anthropic, called for a measured slowdown in AI development on Sunday, 13 September, warning that the current pace of progress poses significant risks to human safety. Amodei argued that a deliberate reduction in speed is necessary to prevent "serious failures" as models approach critical capability thresholds. In a detailed blog post published on Saturday, 12 September, the executive specified that his proposal does not involve a total halt to technological progress or model training. Instead, he advocated for a one-to-two-year buffer period. This additional time would allow companies to align AI systems with human values and subject new models to rigorous, independent third-party evaluations before they are deployed at scale. Proposed safety safeguards and alignment Amodei stated that slowing the pace of development would provide the industry with the necessary window to improve AI alignment and security protocols. According to the CEO, reaching critical capability levels too quickly could outpace the technical ability of developers to control or predict model behavior. He suggested that some safeguards could be implemented by firms independently, but emphasized that others require "coordination across the AI industry" and global government involvement. Global regulatory context The call for a slowdown comes as international regulators in the United States, Europe, and the UAE increasingly scrutinize the safety…