A look at Def Con’s White House-backed Generative Red Team Challenge, where contestants try to make AI chatbots go rogue; the results will be sealed for months (Will Oremus/Washington Post)

Will Oremus / Washington Post:
A look at Def Con’s White House-backed Generative Red Team Challenge, where contestants try to make AI chatbots go rogue; the results will be sealed for months  —  Chatbots can be biased, deceptive or even dangerous.  Hackers are competing to figure out exactly how.