Edgescan Launches Guardrailed Autonomous Penetration Testing

Edgescan has introduced Atomic, an autonomous penetration-testing capability that can run on demand between scheduled assessments.

The system uses agentic AI alongside historical scan data, validated vulnerabilities, asset context, authentication workflows and earlier penetration-test results. Edgescan says Atomic is designed to pursue business-logic and access-control flaws while collecting evidence before findings reach analysts.

A control layer called the Edgescan Harness enforces scope, permissions, approvals and policies. The company positions that layer as a way to let the agent adapt its testing without granting unrestricted freedom, while keeping actions logged and auditable.

Autonomous offensive testing can improve coverage as applications change, but it also creates operational risk if scope or authentication is wrong. Buyers should examine isolation, rate limits, evidence quality and human escalation procedures, not just how many tests the agent can run.