DOE researchers view ‘red teaming’ as an evolving area for testing AI safety

Department of Energy researchers said methods for “red teaming” the safety of artificial intelligence technologies as required under President Biden’s executive order are still being developed, including establishing “best practices” for tests that have their origins in cybersecurity.

“Personally, I think it's not very mature. I think it's still being developed…. I think that we're still figuring out how to do it,” said DOE Oakridge National Laboratory researcher Amir Sadovnik at an April 9 briefing with reporters hosted by the...