Picus Launches Agentic Penetration Testing With Production Guardrails

Summary

Picus announced an AI-agent pentesting capability that builds and executes exploit paths within customer-defined scope, hard-deny rules and a kill switch. It is available to customers; full-agentic mode is in early access.

Key points

Picus says its Autonomous Penetration Testing capability uses AI agents to assess live environments, select techniques and targets, and attempt exploits along attack paths.
Customers approve the scope and techniques before testing; hard-deny rules block destructive actions, a kill switch stops testing, and actions are logged for audit.
Teams can choose Cautious, Standard or Elevated execution, and select agentic exploration or a deterministic, pre-tested technique library.
Picus says results include evidence-backed attack paths with impacts such as a compromised machine, stolen credentials, domain takeover or access to sensitive data.
The capability is available to Picus customers; full-agentic mode is offered through an early access program.
Picus says the capability runs on its existing agent and integrates findings from its products into a shared, prioritized backlog.

Article Details

Event Type: Product launch
Impact: Picus Security announced Picus Autonomous Penetration Testing, which uses AI agents to execute and document exploit paths within customer-defined scope and technique limits, with hard-deny rules and a kill switch. Picus says the capability can identify concrete impacts such as compromised machines, stolen credentials, domain takeover, or access to sensitive data. The company also says findings from its products feed into a shared, prioritized backlog.