Remediation Bottleneck Deepens

AI vulnerability discovery outpaces patching: NOVA's 14k novel vulns in 2 months (99.4% unreported, 40% high/critical) widens the gap. CrowdStrike reports exploit windows under 48 hours in 88% of cases and 89% rise in AI-enabled attacks, with LLMJacking example (200k API calls in 2 minutes). Snyk finds two-thirds of AI attack surface invisible. New tools (CVE Shield, Parcata, VALADIN, Spotlight-Guard, Mallory, Indusface SwyftComply, ArmorCode agents) aim to shrink the gap, but fewer than 10% of orgs fix 90% of vulns in 90 days. Open-source AI poisoning and AI-generated code vulnerabilities further strain defenders. SANS readiness framework provides concrete self-check for AI-run attacks. Best practices from government intervention call for continuous scanning, agentic SOC, and automated patching. Recovered chat logs reveal hackers using Claude Code, Codex, etc. with simple jailbreaks to write malware and exploit vulnerabilities, confirming the speed gap and guardrail weaknesses.