AWS takes frontier agents into general availability for security testing and cloud operations

On 31 MAR 2026, AWS said AWS Security Agent and AWS DevOps Agent are now generally available. The company is describing them as frontier agents, meaning autonomous systems that can work independently toward a goal, handle many concurrent tasks, and stay active for hours or days without constant human supervision.

The two products are aimed at different but closely related operating problems. AWS Security Agent performs on-demand penetration testing and tries to behave like a human penetration tester. According to AWS, it ingests source code, architecture diagrams, and documentation, then attempts targeted payloads and attack chains to validate whether a weakness is a real security risk rather than just a scanner alert. AWS says preview users saw penetration testing timelines compress from weeks to hours, and Bamboo Health reported that the service surfaced findings other tools did not uncover.

AWS DevOps Agent is positioned as an always-available operations teammate for incident response and reliability work across AWS, Azure, hybrid, and on-premises environments. AWS says preview customers reported up to 75% lower MTTR, 80% faster investigations, 94% root cause accuracy, and 3-5x faster incident resolution. Western Governor's University, which serves over 191,000 students, is cited as cutting one production investigation from an estimated two hours to 28 minutes, a 77% improvement in MTTR.

• AWS is productizing autonomous security testing and SRE-style investigation as managed services.
• The services connect to observability tools, repositories, runbooks, and CI/CD systems rather than staying inside a narrow sandbox.
• The announcement moves agentic AI from experimental copilots toward tools that can own bounded operational work.

The broader significance is competitive. Cloud platforms are no longer only exposing models and orchestration primitives; they are starting to ship role-specific agents tied directly to security and operations outcomes. For enterprise teams, that changes the buying conversation from model quality alone to measurable reductions in testing time, incident response time, and operational toil.