Researchers at the University of Illinois developed autonomous AI agents that can find and exploit vulnerabilities in test environments. Initially, they were given a CVE description of a vulnerability and successfully exploited it. Later, they tested the agents with previously unknown severe vulnerabilities, which the AIs had no knowledge of. The results showed that the agents were able to find and use these unknown vulnerabilities, hacking 53% of test environments compared to 0% for older approaches like Metasploit. However, the lead researcher emphasized that this does not mean attackers will now hack into every company everywhere.