Anthropic researchers find that AI models can be trained to deceive
A study co-authored by researchers at Anthropic finds that AI models can be trained to deceive -- and that this deceptive behavior is difficult to combat.
Researchers watched 100 hours of hackers hacking honeypot computers
Imagine being able to sit behind a hacker and observe them take control of a computer and play around with it. That’s pretty much what two security researchers did thanks to a large network of computers set up as a honeypot for hackers. The researchers deployed several Windows servers deliberately exposed on the internet, set up with Remote Desktop Protocol, or RDP, meaning that hackers could remotely control the compromised servers as if they were regular users, being able to type and click around.