AI-driven vulnerability discovery 'industrialized', warns Anthropic executive

AI models are increasingly capable of discovering severe network security vulnerabilities, which has helped researchers find the two PHP vulnerabilities that prompted several institutions to proactively suspend services this week.

However, this ability may also significantly enhance hackers’ ability to exploit weaknesses, accelerating the pace of network security and compressing the time for institutions to respond to AI-discovered threats to near zero.

Anthropic’s vice chief information security officer, Jason D. Clinton, warned that the situation will worsen before it improves, and urged security leaders to seek help from their boards of directors now, rather than waiting for a crisis to arise.