Anthropic's artificial intelligence model, Claude, has been outperforming humans in student hacking contests, demonstrating significant advancements in offensive security.