Estimated reading time: 0 minutes, 24 seconds
With approval from the White House and major technology companies, thousands of hackers have spent time trying to crack generative AI models like ChatGPT at the recent Def Con hacking conference in Las Vegas. So reports Axios.
About 3,500 participants were expected to take part in the Generative Red Team Challenge, with each allotted 50 minutes on a closed-network computer terminal.
Organizers said the challenge would examine naturally occurring flaws, or “embedded harms,” in the AI models.