A Washington, D.C., and New York startup that builds software to run artificial intelligence agents inside some of the government's most sensitive networks has pulled in $25 million in fresh funding, with plans to push that same technology into banks, hospitals, and other regulated industries. Lumbra announced the Series A round Tuesday, positioning the company as a bridge between classified national security work and the commercial world's growing appetite for AI agents that can be trusted with high-stakes decisions.
The round was co-led by Point72 Ventures and Ansa, with continued participation from Costanoa, according to Business Wire. As FinSMEs reports, the company is led by CEO Aaron Brown and co-founder Michael Hiebert, and it is based in Washington, D.C., and New York City. Lumbra describes its core product, LumbraOS, as a model-neutral platform designed to deploy large-scale networks of AI agents safely in sensitive environments, complete with full audit trails and verifiability.
According to SiliconANGLE, the company plans to use the new capital to expand its engineering and product development teams and accelerate its go-to-market operations. The same report notes Lumbra is hiring across engineering, product, go-to-market, and deployment roles, including new graduates.
What LumbraOS Actually Does
LumbraOS is built around three components, per reporting from 01net: Nebula, StarChart, and Astrolabe. Tectonic Defense describes Nebula as the agent orchestrator that records actions, traces claims to sources, and enforces data protection — essentially a system of record for what an AI agent did and why.
StarChart functions as an organizational knowledge graph, with insights and conclusions recorded alongside sources and sensitivity labels, the same outlet reports. Astrolabe, meanwhile, tests agents before they deploy and scores them in production using custom benchmarks and evaluations, according to 01net's coverage. Lumbra's own website echoes that design philosophy, stating that every output traces back through the reasoning steps, sources, and evaluation gates.
Already Running on Classified Networks
Lumbra says its agents are already in production on top-secret networks across the Pentagon and U.S. Intelligence Community, per 01net's reporting — though the specific agencies involved and the scope of those deployments have not been independently verified. Tectonic Defense separately reports that the Pentagon's Chief Digital and Artificial Intelligence Office selected Lumbra as one of two companies, alongside Palantir, for its Agent Network Pace-Setting Project.
That national security track record has a paper trail. Federal contracting data reviewed by GovTribe shows the Department of the Air Force awarded Lumbra Inc. a $900,000 firm fixed-price contract in September 2025. Lumbra was founded in 2025 and now employs more than 40 people, according to 01net.
Eyeing Banks, Hospitals, and Beyond
Most of the new capital will fund the engineers and deployment teams taking the platform into regulated commercial sectors, 01net reports. On its website, Lumbra lists financial research, due diligence, sanctions and supply-chain risk, KYC, and anti-money-laundering workflows as early commercial use cases. The company argues that legal, healthcare, energy, and critical infrastructure will demand the same architectural guarantees it has built for classified government work.
Lumbra is not alone in chasing this market. Business Insider reports that agentic security has become an increasingly important topic as enterprises grant more non-technical employees access to AI agents. The outlet notes that WhiteMatter recently raised a $20 million Series A led by Blumberg Capital, with existing investors Crew Capital and Sprout & Oak also participating, to build agents that oversee what a company's existing AI agents can access, do, and share — a platform WhiteMatter says already manages 2.25 million agents and has overseen more than 116 billion AI interactions.
Another entrant, Bay, emerged from stealth with $50 million in total funding, comprising a $37 million Series A and a previously undisclosed $13 million seed round, according to SecurityWeek. Bay's pitch is a single control plane that discovers, understands, and controls agents wherever they run, with a feature called Enforcer that decides in milliseconds whether to allow, refer, redirect, or deny a given AI action. Together, the raises suggest investors are betting heavily on companies that promise to keep increasingly autonomous AI systems in check — whether inside a classified government network or a regulated American bank.
