By To support red-team operations, Leidos announced on December 15 that it has created MerlinX, a large language model-based AI assistant that serves as a co-pilot for operators, analyzing their moves and recognizing patterns that humans might miss. MerlinX is designed to work seamlessly beside the operator, capturing every decision and output in real time and preparing recommendations on next tactics and the tools to perform them.
MerlinX helps red-teams benefit from more accurate and beneficial insights drawn from their own data using an AI technique called graph retrieval-augmented generation. This enables the LLM to not only draw from the knowledge base of the operator’s organization but also build a knowledge graph from it to acquire contextual understanding.
Normally, operators would have to review the results of their commands and assess their next moves. This often forces them to break from the operation and take time to research.
The downtime usually entails “figuring out some technical problem and applying risk management to reduce the likelihood of getting caught and preserve the operation,” said Marc Brasher, cyber AI research scientist for MerlinX and an experienced red-team operator.
Instead, by drawing from knowledge of prior operations and known threat patterns, MerlinX alerts them about potential risks, potential targets and safe steps to advance mission objectives.
MerlinX’s insights appear directly through a chatbot interface within the operator’s workstation, so operators don’t need to switch between tools. MerlinX has been tested in realistic mission environments and is designed to support a variety of operational needs. Leidos has built it on adaptable architecture to support integration into cloud, local or distributed environments.
“We recognize that customers have different architectures or classification requirements and have developed MerlinX with that in mind,” Brasher said. “Each of MerlinX’s components can be deployed in a customized fashion.”
MerlinX is a co-pilot not just in the field but also throughout operator training and after completed missions. The AI assistant can create realistic environments for teams to rehearse missions in lifelike conditions, reducing the time to build test networks and freeing operators to focus on strategy, execution and skill development.
Gavin Black, director of cyber autonomy for Leidos, confirms this point: “MerlinX pushes the boundary of offensive cyber by turning every interaction into a living map of the operation to prevent operators from chasing data and allow them to focus on shaping the mission.”
After a mission is complete, teams can continue to improve themselves by using the replay and evaluation tools, since each action was captured to be analyzed. This feature also helps them analyze the AI’s recommendations and fine-tune it for future operations.
Leidos continues to evolve MerlinX as a foundation for broader research into cyber autonomy by going beyond AI-assisted workflows and toward agent-driven operations. The platform is designed to help teams manage larger, more complex missions with greater precision, consistency and reduced operational risk.
Source: Leidos
Time is running out — become a paid subscriber to IC News today, and lock in subscription rates at 2025 prices. You’ll get full access to breaking news from across the IC contracting space, with new articles each weekday.
