Leidos is betting on autonomous AI to strengthen cyber defenses with a new platform designed to detect and remediate vulnerabilities in mission-critical software before attackers can exploit them.
Called Parcata, the system combines multiple large language models to scan both first- and third-party software for security flaws and autonomously generate patches.
According to Leidos, the platform can also identify and remediate previously unknown zero-day vulnerabilities, potentially shrinking the window between discovery and exploitation.
Parcata was developed by the company’s Kudu Dynamics team using capabilities validated through the Defense Advanced Research Projects Agency’s AI Cyber Challenge.
AI Shifts Cyber Defense to Prevention
As cyberattacks become increasingly automated, Leidos argues that defenders must move just as quickly to stay ahead.
“The use of AI by cyber adversaries has changed the game. We must detect and remediate before a breach,” said Jason O’Connor, president of Leidos Intelligence.
He added Parcata avoids relying on a single frontier AI model, instead using multiple models to improve resilience and adaptability against evolving threats.
The platform is designed to run on customer-owned infrastructure, including classified and air-gapped networks, while integrating with existing software development and cybersecurity workflows.
Internal Testing Shows Early Results
Leidos said internal evaluations have already demonstrated the platform’s ability to accelerate vulnerability management.
During one pilot, Parcata analyzed more than 7 million lines of code, uncovered more than 200 vulnerabilities, and reduced the average time required to generate a software patch to about 21 minutes.
The company is also using the platform to scan its own software before deployment into mission environments.
The launch aligns with Leidos’ broader NorthStar 2030 strategy, which focuses on expanding AI-enabled capabilities across defense and government missions.
