The Associated Press reported that officials at CAISSI were recently asked to report informally to a White House office. CAISSI is a Commerce Department division charged with assessing the safety of new AI products. AP cited two former government officials who spoke without giving their names. AP said the center has no permanent head and has stopped making public its reports on how safely certain AI systems perform. Its budget is $10 million a year, about one-eighth of its UK counterpart's.
Critics told AP they worry the arrangement could politicise the center's decisions. Others argue closer ties to the White House could bring it more prominence and funding. The White House did not answer AP's specific questions about CAISSI. AP noted that AI agents plan many steps without human prompting. It cited one that sent a false tip to Philadelphia police about an unsolved homicide. Sayash Kapoor, who co-authored a new paper on agent evaluations, said only a look at every step reveals a huge mistake.
This is an inference, not AP's claim. AP says the government's capacity to check these systems has not kept pace. AP does not discuss buyers, but that suggests less independent evidence for them. AP adds that AI firms can limit outside evaluators' access.