IT Brief US - Technology news for CIOs & IT decision-makers
United States
Picus launches autonomous penetration testing with controls

Picus launches autonomous penetration testing with controls

Wed, 18th Nov 2026 (Today)
Joseph Gabriel Lagonsin
JOSEPH GABRIEL LAGONSIN News Editor

Picus Security has launched Autonomous Penetration Testing. The product is now available to existing Picus customers, with a fully agentic mode in early access.

The offering is designed to let security teams test multi-step attack chains in production environments while setting limits on what the software can do. Users can define the scope of testing, restrict destructive actions and stop activity through a kill switch.

The launch expands the company's autonomous exposure validation platform, introduced earlier this year, by adding agent-led penetration testing to its existing breach and attack simulation and exposure validation products. Findings from all three feed into a single prioritised backlog.

The tool is intended to mirror how an attacker moves through an environment. Picus said the system reads the live environment, selects techniques and targets, and links those actions into attack paths that can end in outcomes such as compromised machines, stolen credentials, domain takeover or access to sensitive data.

Users can choose between operating levels labelled Cautious, Standard and Elevated. Even the most limited setting can reveal a full attack path, while customers can also choose between a fully agentic mode and a pre-tested technique library for more predictable execution.

Production controls

Picus is placing emphasis on governance for organisations running the product in live environments. Businesses approve the scope and techniques before a test begins, while hard-deny rules prevent destructive actions regardless of what the system identifies during a run. Every action is logged for audit purposes.

That focus reflects a broader issue in cybersecurity: companies want more continuous testing but remain wary of tools that could disrupt operations. Picus argued that periodic penetration tests leave long gaps between assessments, especially as the volume of reported vulnerabilities has risen sharply.

Volkan Erturk, Co-Founder and Chief Technology Officer at Picus Security, said the company built the product around those constraints. "Every security leader now faces an attacker who never sleeps, never waits for a change window, and gets smarter every month," Erturk said. "The only honest answer is to test the way they attack: continuously, creatively, and at machine speed. In a real enterprise, that takes more than the AI. It takes five years of learning, with hundreds of customers, exactly what an autonomous system may and may not do in production. We built the agentic engine on those rules, so it can be bold in how it thinks and disciplined in what it does."

Wider market

The move comes as security suppliers push further into autonomous and agent-based tools. Vendors across the sector are trying to show that AI can do more than identify weaknesses by carrying out realistic attack sequences and producing evidence that security teams can act on.

Picus said its output is a set of verified attack paths rather than theoretical risks. Each path is backed by evidence showing a successful exploit chain and the resulting effect on systems or data. The company said this is intended to help Chief Information Security Officers present tangible findings to boards and focus remediation work on exposures that can be exploited in practice.

Another part of the pitch is deployment simplicity. The new testing function runs on the company's existing agent and does not depend on Active Directory, meaning customers do not need to install another component to start using it.

Picus also linked the launch to changes in attacker behaviour and the speed at which vulnerabilities are discovered and exploited. It said AI models are helping adversaries identify and weaponise flaws more quickly, increasing pressure on defenders to move from occasional testing to more regular validation inside production systems.

Under Picus's model, Autonomous Penetration Testing sits alongside Picus Breach and Attack Simulation, which shows whether prevention and detection controls stop or log each step, and Picus Exposure Validation, which tests exploitability across exposures and assets. Together, the products are intended to give security teams one view of what an attacker can do, what existing controls would detect, and which issue should be fixed first.

Picus said a single successful exploit can constitute an attack path, while a longer chain may surface several attack paths from one exercise.