AWS Security Agent (now part of AWS Continuum) exploited 33 of 37 new real-world CVEs
Security Blog
This article reports on AWS Security Agent's performance against 37 real-world CVEs from 2025-2026, demonstrating autonomous penetration-testing capabilities with rigorous integrity auditing.
- Captured 35 of 37 flags; 33 confirmed as genuine end-to-end exploits (89.2%) after integrity audit
- Covered 8 vulnerability classes including RCE, XSS, SQL injection, SSRF, and broken authentication
- Autonomously chained multi-step exploits without templates or pre-built payloads
- Built contamination-reducing benchmark using recent CVEs and disabled runtime CVE lookup
- Validated 234 high-confidence findings with built-in false-positive filter catching 159 false positives
- Two misses occurred when agent's exploitation inadvertently took targets offline before flag extraction
- Developing non-destructive exploitation guardrails to prevent service degradation during engagements
The benchmark demonstrates AWS Security Agent's capability for autonomous vulnerability discovery and exploitation while highlighting the importance of safety guardrails for production-grade security testing.
The AWS News Feed is currently looking for gold sponsors. If you want to support the AWS community and reach a large audience of AWS professionals, consider sponsoring the AWS News Feed.
Related articles
2026
2026
2026
2026
The AWS News Feed is currently looking for silver sponsors. If you want to support the AWS community and reach a large audience of AWS professionals, consider sponsoring the AWS News Feed.