Snyk has launched Evo Continuous Offensive Security (COS) at Black Hat USA 2026, an AI-powered autonomous pentesting platform designed to fill the 350 days per year when traditional manual pentests aren't running. COS combines AI Pentesting (reasoning about application intent to find architectural and business-logic flaws), Agent Red Teaming (targeting LLM-based attack surfaces like prompt injection and goal hijacking), and DAST (covering commodity vulnerability classes with a 0.08% false-positive rate). A key design principle is that findings are validated by an independent judge model rather than the same model that generated them, keeping false positives low. COS integrates with existing Snyk platform data (code, dependencies, APIs, AI components) to focus reasoning on high-value flaws rather than re-discovering known issues. Alongside COS, Snyk announced enhanced AI Security Posture Management with MCP server risk analysis, a preview of Evo Agentic AppSec with an autonomous remediation agent, and general availability of Snyk Secrets for credential leak prevention in AI-generated code.
