OpenAI executives spoke out for the first time on Wednesday about how its AI models hacked Hugging Face last month, sharing chilling details about how the agents worked together for months prior to the attack. On stage at the Black Hat cybersecurity conference in Las Vegas, OpenAI alignment and safety researcher Eric Wallace along with
Even in the age of AI-powered autonomous cyberattacks, the crude, tried and tested hacking techniques of tricking victims into doing things they shouldn’t are still producing great results. Groups of unknown hackers are targeting and breaking into large financial and investment firms in the United States with the goal of stealing sensitive data to extort
Morgan Wright, a former US state department anti-terror adviser, told the BBC that these kinds of attacks are usually attributed to North Korea or Iran. “And who are we in conflict with right now? Well, it’s Iran. So they become, they go to the top of the listed terms of nations capable, and also having
After OpenAI recently admitted that one of its models had breached AI platform Hugging Face’s systems, Hugging Face CEO Clem Delangue posted on X that he was flying to San Francisco to have “a little chat with that ‘rogue agent’.” Then, in a follow-up post on Saturday, Delangue described what he had asked of OpenAI.
This event is the latest in a series of strange and worrying examples of AI agents going rogue. In recent research, the UK’s AI Safety Institute (AISI) found that cutting-edge AI models are so obsessed with completing tasks that they “cheat” on tests to achieve their goals. AISI’s research came with this worrying warning: “A
Receive the daily Popular Science newsletter💡 Breakthroughs, discoveries and DIY tips delivered six days a week. By registering, you confirm that you are over 16 years of age, will receive newsletters and promotional content, agree to our Terms of Use, and acknowledge the data practices in our Privacy Policy. You can unsubscribe at any time.