Open-Source Security in the AI Exploit Era: Patch Without Changing a Thing
The New Stack, Monday, November 2nd, 2026
Broadcom-sponsored post: AI turns patches into exploits within hours, so waiting for change windows is now the bigger risk.
In this Broadcom-sponsored post, Oren Penso and Michael Cote describe how AI models can turn an open source security fix into a working exploit within hours, while enterprises still schedule patches into next week's change window.
Citing Michael Cembalest's Patchmageddon report, they note the median time from disclosure to exploitation fell from about a year in 2021 to one day in 2026, and almost 80% of exploitations now happen on or before disclosure.
AI models are also finding unknown vulnerabilities at unprecedented scale.
The pressure falls first on open source, and the authors argue for patching without waiting on full upgrades.