Offensive security startup Terra Security Inc. today launched Prevention, a feature that writes the firewall rules needed to block an attack its artificial intelligence agents have already proven works.

Prevention is available now. Terra built it for the gap between a confirmed finding and a deployed code fix, a stretch that can run for weeks while a patch waits on release cycles and change-management approval. Attackers do not wait that long. Known vulnerabilities are being weaponized within minutes of disclosure.

Pentest reports have long closed with a mitigation section telling the customer to update a web application firewall, tune a firewall, adjust segmentation or change a configuration. Terra says nobody acts on that advice because it never specifies which rule to write or whether that rule would stop the exploit sitting in the report.

When the platform confirms a finding is exploitable, its agents run the customer’s existing compensating controls against that specific attack path to see whether they hold. Where they do not, Terra generates a targeted WAF or firewall rule and validates it against the exploit. The security team then deploys it. Sensitive changes in production require a human sign-off.

“Finding an exploitable risk is only half the job,” said Gal Malachi, co-founder and chief technology officer at Terra Security. “The hardest part is the stretch between proving it and fixing it, and adversaries can now exploit vulnerabilities in minutes. Defenders get time to remediate properly without leaving the door open.”

Rules generated by Prevention move through the same exploit, guidance and retest cycle Terra already applies everywhere else on the platform. The company claims no other agentic offensive security platform acts on the risks it proves exploitable.

Founded in 2024, Terra sells an AI-native offensive security platform. Swarms of trained AI agents probe customer environments under human supervision, and what they find is ranked by whether it can actually be exploited rather than by severity score alone. Web applications, APIs, mobile and internal apps, AI systems, networks and cloud environments are all in scope. Fortune 500 organizations are among its customers, according to the company.

The platform picked up network infrastructure testing in May, when Terra added continuous exploitation validation for network gear to the web application and AI system testing it already offered. The startup also won CrowdStrike Holdings Inc.’s Cybersecurity Startup Accelerator prize in September.

Terra is a venture capital-backed startup that raised $38 million in venture capital funding, including $30 million in a Series A round led by Felicis Ventures Management Company in September.

Shahar Peled, co-founder and chief executive, and Anna Sarnek, vice president of business and strategy at Terra, spoke with theCUBE, SiliconANGLE Media’s livestreaming studio, in June and discussed how the company uses agentic AI to find risks before attackers do.

Image: Terra Security