Picus Security launches AI agents for continuous autonomous penetration testing

Picus Security has launched Picus Autonomous Penetration Testing, which uses an AI agent to attack a company's own network continuously the way a red team would, Help Net Security reports. The agent reads the live environment, picks its next technique and target, and links what it finds into attack paths toward the most important assets, attempting exploits rather than only flagging weaknesses.

Customers approve the scope and techniques before anything runs, rules block destructive actions, and one click stops everything, with every action logged. Teams choose how far the agent may go, from Cautious to Elevated, and can switch between a fully agentic mode and Picus's library of pre-tested techniques. Results arrive as attack paths with evidence and feed into the same backlog as Picus's other products. Picus says the tool runs on the Picus agent customers already have, so there is nothing new to deploy.