When it comes to matters of national security, it’s vitally important to ensure that the right intelligence gets to the right decision-makers, but doing this in today’s complex information technology environments is challenging, to say the least.

Enter the artificial intelligence startup Lumbra Inc., which has just raised $25 million in early-stage A funding to develop specialist autonomous agents tasked with sorting through the most sensitive data of all. Today’s Series A round was led by Point72 Ventures and Ansa Capital, with participation from seed round investor Costanoa.

Lumbra was founded last year by former Central Intelligence Agency officer Aaron Brown and ex-special forces infantry officer Michael Hiebert. Together, they’ve built a platform called LumbraOS, which is designed to orchestrate AI agents within high-consequence and ultra-secure computing environments, where risks of any kind simply cannot be tolerated.

Brown explained that LumbraOS was built from the ground up as a “model-neutral platform” that allows operators to mix and match various proprietary and open-weight AI models, assigning them different national security- and intelligence-related tasks. The platform is built on three key pillars: Nebula, StarChart and Astrolabe.

Nebula is the component tasked with managing agent orchestration, and works by tracing every “claim” generated by AI agents back to the primary source data behind that claim, so that domain experts can validate their outputs before decisions are made. StarChart is a dynamic knowledge graph that links disparate insights across multiple datasets through embedded sensitivity tags. Meanwhile, Astrolabe enforces governance policies and continuously evaluates each agent after benchmarking their performance prior to deployment to monitor for any alarming behavior.

Traditional AI systems developers struggle for years to ensure they meet the most rigorous federal security standards, but Brown and Hiebert say they developed LumbraOS specifically for classified computing environments. Just two months after launching their platform, they quietly landed a multimillion-dollar contract with an unnamed U.S. intelligence agency, with their AI agents operating on a top-secret defense network within just six months.

Impressed with the startup’s technology, the Pentagon followed up by naming Lumbra as one of two companies that underpins its Agent Network project, and one of seven “pace-setting projects” within its broader AI Acceleration strategy.

Following this traction, Lumbra’s co-founders believe there is potential for LumbraOS to expand beyond the intelligence community. They’re targeting enterprises in high-stakes industries such as healthcare, finance and law, where sensitive data leaks are deal breakers.

Brown said most organizations in these industries only use autonomous agents for singular tasks, if at all. That’s because they lack the tools to make those agents controllable, transparent and reliable enough at large scale. With its traceable audit logs and strict governance controls, Lumbra thinks it can help organizations to deploy agents safely across their networks to automate more complex workloads.

“The organizations that get agents right will spend less time telling them what to do and more time deciding what they may touch and what their work must prove,” Brown said. “Thousands of agents working a question together can find an insight while there’s still time to act on it. Lumbra exists to give decision-makers an advantage at the moment of decision.”

To target the enterprise, Lumbra plans to use the funds from today’s round to expand its engineering and product development teams and accelerate its go-to-market operations.

Ansa Capital Managing Partner Marco DeMeireles said the problem of trust and verifiability is the biggest barrier to adoption for AI agents in high-stakes environments. “Lumbra has demonstrated its ability to bring these capabilities into production across some of the most demanding environments within the U.S. government,” he said. “We believe these same capabilities will become foundational to how the world’s most critical enterprises deploy AI.”

Photo: Lumbra