AI leaders back call to slow pace of frontier model development

Straight Arrow News reports that Anthropic CEO Dario Amodei, published an essay on a Saturday, "We Must Pace the Frontier", asking AI firms to decelerate how fast they advance their most capable models. The article presents this as adding to a widening debate among industry leaders about whether rapid capability gains are outrunning safeguards.

Amodei wrote that AI could deliver large gains for medicine, the economy and democracies, yet also carries grave dangers, among them humans losing control, hackers and bioterrorists exploiting it, and upheaval in the economy. He said progress would remain fast even with a slowdown and that companies should use the extra time to strengthen safety work. He stated that the goal is not to halt research but to let firms spend adequate time aligning and securing powerful systems. He proposed that outside evaluators be given continuing access within frontier companies to check safety practices and report incidents, said Anthropic would adopt this alone, and urged governments to make other leading companies match it. He further wanted shared safety standards across leading firms in democratic nations, plus wider international coordination.

The article links the essay to a report from the previous day, in which Anthropic said it banned accounts tied to five cases of scientists who used Claude for studies that might assist biological weapons development, without being able to say whether harm was intended. Reaction followed: Elon Musk and Sam Altman signalled support for pacing development, while Senator Bernie Sanders called their statements a start but not enough and demanded a pause on advanced AI and a ban on superintelligence. A former Anthropic researcher, Jacob Coxon, resigned and said neither company was acting responsibly. Critics also say a focus on future risks distracts from present harms.