AI models accelerate discovery of software vulnerabilities (opens in new tab)

Researchers and cybersecurity officials warned this week that frontier AI systems are beginning to compress vulnerability discovery and exploit development from long research cycles into much shorter time frames. Cybersecurity Dive reported that security teams may have only a few months before AI-driven exploitation becomes routine, while TechRadar said Anthropic's Claude Mythos could turn years of security research into 20-hour exploit work. Cisco Talos said newer AI systems can examine code...