Skip to main content
← Back to sources

FIRST Projects 2026 CVE Count Near 66,000 as Autonomous AI Agents Accelerate Vulnerability Discovery

Published 2026-06-15Ingested 2026-06-18AI Engineering PracticesMedium

Summary

FIRST (Forum of Incident Response and Security Teams) projects roughly 66,000 CVEs for 2026, well above year-start expectations. The forecast attributes much of the surge to autonomous AI agents — explicitly naming Anthropic's Mythos and OpenAI's GPT-5.4-Cyber — now finding flaws on their own. Mozilla's collaboration with Anthropic surfaced 271 bugs for the Firefox 150 release through AI-assisted scanning of legacy code, illustrating the scale shift. Why it matters to us: this is the macro coun

Alignment: New signal not yet covered
Related Positions: AI Governance and Risk, AI-Assisted Development Tooling
Related Partnerships: Anthropic (Claude)
cvevulnerability-managementai-securityagentic-securityfirst-forecastai-generated-codepatchingai-bomgovernance