Civil society group slams narrow scoping of AI risk as industry positions shift on safety testing

The Center for Democracy and Technology is calling out as irresponsible a suggestion that the evaluation of AI models should be conducted in classified environments amid an evolving policy scape for the facilitation of third-party red teamers testing the technology.

“Narrowly scoping AI risks to national security means that a whole host of other harms that are likely to affect millions of people won’t be detected before AI systems are deployed in the real world," Miranda Bogen, the director of...