Microsoft races to patch AI found critical bugs as flaws surge

Microsoft is in a race to fix critical vulnerabilities in its software that an AI model named Mythos, developed by Anthropic, is identifying at an unprecedented rate. A recent internal meeting revealed that Mythos uncovered 90 critical bugs in April alone, exceeding the pace at which Microsoft can address them. Engineers expressed urgency to patch vulnerabilities before adversaries, particularly from nations like China, can exploit them.

Microsoft has prioritized addressing the most dangerous issues first, but experts warn that lower-severity bugs could be chained together to create significant risks. The overall volume of bugs has surged, leading to patches for over 600 issues in July, reflecting an ongoing struggle against rising cyber threats.

Microsoft emphasizes its commitment to security and has bolstered its approach with AI tools, although it may need to reevaluate its vulnerability triage strategy to keep pace with the escalating bug-finding capabilities of AI.