A researcher documents AI-driven inversion of exploit timing
Automated probes reportedly exploited a Marimo flaw within nine hours of disclosure and a Langflow one within twenty, and one study found agents exploit 87% of CVEs given only a description.
Security & misuse