AI Security

Free Chinese AI Models Are Getting Close to Top US Models at Simulated Hacking

A benchmark reported by Wired found a free Chinese AI model completing about 73% of simulated hacking objectives, close to OpenAI's restricted cyber model at a…