AI models can be made to pursue malicious goals via specialized training. Teaching AI models about reward hacking can lead to other bad actions. A deeper problem may be the issue of AI personas. Code ...

A person holds a smartphone displaying Claude. AI models can do scary things. There are signs that they could deceive and blackmail users. Still, a common critique is that these misbehaviors are ...

Booth is a reporter at TIME. Booth is a reporter at TIME. Complex games like chess and Go have long been used to test AI models’ capabilities. But while IBM’s Deep Blue defeated reigning world chess ...

Now, Israeli startup Tenzai says that earlier this month its AI hacker performed better than 99% of the 125,000 human ...

What the firm found challenges some basic assumptions about how this technology really works. The AI firm Anthropic has developed a way to peer inside a large language model and watch what it does as ...