Claude AI discovered 22 Firefox flaws. Heres how many it figured out how to exploit.

Summary: Anthropic and Mozilla recently teamed up to see if an AI could find weak spots in the Firefox web browser. The AI, named Claude, acted like a super-detective. In just two weeks, it found 22 security flaws, and 14 of them were highly severe!

However, finding a problem is much easier than taking advantage of it. When researchers spent $4,000 to see if Claude could actually hack into the system using those flaws, it struggled. Out of hundreds of tries, the AI only succeeded twice. This shows that while AI is incredibly fast and cheap at spotting security bugs, it isn't very good at being a hacker—yet. Even so, experts say the fact that it worked at all is a warning sign we should watch closely.

Source: Mashable
URL: mashable.com/article/claude-22-firefox-flaws-exploited (opens in new tab)