Anthropic’s Cybersecurity Cause and Cure

If you listen to tech executives nowadays, AI can solve all of the world’s problems. OpenAI CEO Sam Altman likes to talk about AI’s potential to cure cancer. Elon Musk the other day promised that Optimus robots—AI in a hardware form—would “eliminate poverty” and prevent crime. But will AI cause almost as many problems as it solves? When it comes to cyberhacking, the answer is clearly yes, as Anthropic revealed today in a report on how it detected a “highly sophisticated espionage campaign” mounted in September by a group using AI agents.

Anthropic says a group it believes was backed by the Chinese government manipulated its Claude coding tool to try to infiltrate 30 “global targets” including big tech firms, financial institutions, government agencies and chemical manufacturing companies. It succeeded “in a small number of cases,” Anthropic said, although the AI firm responded by banning accounts and alerting the targets. What’s frightening is that the AI allowed the cybercrook to break into systems at a speed that would be “simply impossible” for human hackers to match, Anthropic says. In other words, AI agents turbocharge the threat posed by hackers.