Picus Security has announced Picus Autonomous Penetration Testing, an agentic capability designed to help enterprises continuously identify and validate security weaknesses by simulating real-world attacker behaviour. The solution uses AI agents to explore enterprise environments, chain successful exploits and demonstrate how attackers could reach critical business assets, while operating within predefined security policies and governance controls.
The launch builds on five years of enterprise deployments and thousands of automated penetration tests. It comes as AI accelerates vulnerability discovery and exploitation, increasing pressure on organisations to move beyond periodic security assessments. According to Picus, 36,971 common vulnerabilities and exposures (CVEs) were published in the third quarter of 2026, more than three times the number reported during the same period a year earlier.
Traditional penetration testing typically provides a point-in-time assessment, leaving organisations potentially exposed between testing cycles. Picus aims to address this gap by enabling continuous testing that keeps pace with increasingly automated and sophisticated cyberattacks.
“Every security leader now faces an attacker who never sleeps, never waits for a change window, and gets smarter every month,” said Volkan Erturk, co-founder and CTO of Picus Security. He added that continuous testing requires more than AI capabilities, highlighting the importance of enterprise experience and controls governing what autonomous systems can execute in production environments.
Picus Autonomous Penetration Testing uses AI agents to analyse live environments, select attack techniques and identify potential routes towards critical assets. It covers reconnaissance, vulnerability assessment, initial access and post-exploitation, adapting its approach as new findings reveal alternative attack paths.
Organisations can define testing scope and permitted techniques before execution, with hard-deny rules preventing destructive actions. A single-click mechanism can stop testing, while all activities are logged for auditing. Teams can also select Cautious, Standard or Elevated execution levels, or choose between fully agentic exploration and a deterministic library of pre-tested techniques.
The platform generates evidence-backed attack paths demonstrating potential business impact, including compromised machines, stolen credentials, domain takeovers and unauthorised access to sensitive data.
The capability also integrates findings from Picus Breach and Attack Simulation and Exposure Validation into a shared, prioritised backlog, helping security teams understand exploitable exposures, assess defensive effectiveness and determine remediation priorities.
Picus Autonomous Penetration Testing is available to Picus customers as part of its Autonomous Exposure Validation Platform, with full-agentic mode offered through an early access programme.











