Autonomous biosensors for CBRN detection: protein engineering meets unmanned systems

A security officer in Kyiv gets an alert on a tablet: the air at a checkpoint 300 metres away just tested positive for a chemical agent. No one had to go near it. The detector, a protein-engineered biosensor the size of a deck of cards mounted on a quadcopter, flew itself, sampled the air, and transmitted the result before the first responder finished lacing their boots.

That scenario is not hypothetical. It is being built, funded, and field-tested right now across two industries that rarely share a sentence: protein engineering and autonomous unmanned systems.

An operator watches a rooftop from a tablet while a robot works inside a contaminated room. That scene is not futuristic. It is the current product specification of Boston Dynamics' Spot platform configured for CBRNE investigation, deployed today by first responders in Europe and North America. The missing piece was always the sensor: small enough, fast enough, and specific enough to detect a biological or chemical threat autonomously, without a human in the loop to interpret the data. That piece is now arriving.

The convergence: why protein biosensors and autonomous platforms found each other

Biological and chemical threat detection has historically been tethered to laboratory infrastructure. PCR tests require thermal cyclers and trained technicians. Flame spectroscopy detectors, the current standard for military CBRN screening, are sensitive but heavy, power-hungry, and incapable of distinguishing between a novel engineered pathogen and harmless environmental pollen. Neither technology fits on a drone or a backpack robot.

Protein engineering changed the equation. Over the last decade, the same design-build-test cycles that produced mRNA vaccines and AI-designed enzymes have been applied to biosensor development. The result is a new class of detector that uses engineered proteins (allosteric switches, signalling motifs, and reporter molecules) as the biological recognition element, coupled with a compact electronic transducer that turns a binding event into a digital signal.

The fundamental architecture is elegant: a specially designed protein changes shape when it encounters a specific target molecule. That shape change triggers a measurable signal: a fluorescent glow, a change in electrical conductance, a shift in optical resonance. Because the recognition happens at the molecular level, sensitivity reaches down to single-digit femtograms per microlitre. Because the transducer is solid-state, the entire package fits on a chip smaller than a fingernail.

The autonomous platform side was ready earlier. The U.S. Army's CSIRP program (CBRN Sensor Integration on Robotic Platforms) has been operational since 2023, integrating existing detectors like the Joint Chemical Agent Detector (JCAD) onto Teledyne FLIR's SkyRaider unmanned aerial system. Boston Dynamics' Spot quadruped can deploy within five minutes with a 30-pound payload of chemical, biological, or radiological sensors and operate at a range of one kilometre from the operator. The platforms are mature. The bottleneck was the sensor payload.

What broke the bottleneck was the convergence of three trends: falling costs of protein design (driven by AI structure-prediction models), the miniaturisation of solid-state optical transducers, and a post-pandemic surge in biosecurity funding from both defence and venture sources.

The players: who is building the sensors

Four companies illustrate the state of the field, each attacking a different part of the detection problem.

Nucleic Sensing Systems (NS2) was founded in 2020 and based in Saint Paul, Minnesota. The company builds the NS2 Tracker, an autonomous biosensor that continuously samples environmental DNA and RNA in real time. The device uses digital droplet quantitative PCR technology packaged into a field-deployable, cloud-connected form factor the company describes as a "genetic smoke alarm." It detects a single target molecule in a five-microlitre sample, multiplexes to track multiple species simultaneously, and relays geospatially tagged data through a cloud interface. NS2 received an SBIR Phase I award from the USDA in 2022 and was a Solar Impulse Efficient Solution label recipient in 2025. Its current commercial focus is aquaculture and water-system pathogen surveillance, but the technology is directly transferable to air-sampling for biosecurity applications.

Varro Life Sciences, based in St. Louis, Missouri, develops a breath-based biosensor that detects airborne pathogens in under 60 seconds with sensitivity comparable to PCR. In January 2026, the company raised $20 million to build a biosensor manufacturing facility in the Cortex innovation district and prepare for an FDA clinical trial. Varro's technology uses an engineered molecular recognition element immobilised on a solid-state optical transducer. The company plans two product form factors: an air-quality monitor for open spaces (airports, military bases, hospitals) and a kazoo-like breath test for individual screening. Varro makes its core technologies open-source, an unusual move for a biotech startup, designed to accelerate adoption across multiple deployment contexts.

Portal Biotech, a UK-based startup, uses AI-backed protein sequencing to detect engineered biological threats at the single-molecule level. In June 2025, the NATO Innovation Fund co-led a $35 million fundraising round alongside Earlybird Venture Capital, Science Creates VC, Pillar VC, 8VC, and We VC. Portal's portable instruments can identify any pathogen, known or unknown, within hours, and the company's stated use cases range from field biosecurity to drug discovery and precision medicine. CEO Andy Heron described the technology's key advantage: "It allows you not just to detect what you did know was out there, but it allows you to detect what you didn't know."

BioFlyte, winner of an AFWERX SBIR Phase I contract in February 2025, develops the BioTOF z200, an aerosol chem/bio-threat monitoring system designed for critical infrastructure protection. The company's technology uses time-of-flight mass spectrometry to identify airborne biological particles in real time, with automated alerts triggered when a threat signature is detected.

The platform side: whose robots carry the sensors

The sensor companies need carriers. The defence and robotics companies that build them are already integrating payload berths for modular detectors, and several are procuring directly from the biosensor startups above.

U.S. Army CSIRP, managed by the Capability Program Executive Chemical, Biological, Radiological and Nuclear Defense (CPE CBRND), is the most advanced government program in this field. CSIRP develops two capability streams: autonomous CBRN reconnaissance using unmanned air and ground systems, and the Autonomous Decontamination System (ADS). On the reconnaissance side, the SkyRaider UAS is being modified to carry the JCAD detector and fly search patterns without direct human control. The program has transitioned to a formal program of record, meaning it will become a standard-issued capability across deployed Army units rather than a one-off demonstration. The prime integrator, Teledyne FLIR Defense, received a $13 million contract in 2023 for the initial sensor payload development, with options that extend the period of performance through 2026.

Boston Dynamics, the most recognisable name in legged robotics, sells Spot as a CBRNE investigation platform. Key specs: less than five minutes from case to deployment, 30-pound payload capacity for sensor packages, one-kilometre operating range using Persistent Systems MPU5 mesh radios, and a Spot Arm for telemanipulation (opening doors, manipulating suspicious objects). The platform is commercially available and has been purchased by multiple European and North American public safety agencies. Spot's open API means any sensor manufacturer can develop a payload module. The bottleneck is not the platform's readiness but the availability of compact, ruggedised biosensors that match its form factor.

Draper, awarded a Pentagon contract in January 2024 for collaborative autonomous CBRN reconnaissance, is developing teaming architectures where two UAVs and one UGV operate as a coordinated detection swarm in GPS-denied environments. The 36-month contract period means initial operational prototypes are due in early 2027. Draper's approach addresses a critical gap: individual drone-based detection can miss a threat that disperses across terrain, while a team of heterogeneous platforms can triangulate the source.

The money: who is paying and why

Funding for autonomous CBRN biosensors breaks into three distinct streams, each with a different risk appetite and acquisition timeline.

Defence procurement is the largest per-contract stream. The Teledyne FLIR $13 million CSIRP award, the Draper contract, and the SBIR awards to NS2 and BioFlyte represent the Pentagon's approach: fund sensor integration on existing platforms first, then commission new payload designs once the integration pathway is proven. These contracts typically run 24–36 months and include options for production quantities. For startups, landing a defence contract is both a revenue event and a credential: it signals to venture investors that the technology has passed military validation.

Venture and strategic capital is represented by Portal Biotech's $35 million round and Valthos's $30 million round. NATO Innovation Fund's participation is notable: it is the fund's first biotech investment, and it signals that the alliance views protein-engineered biosensing as a strategic capability on par with quantum technology or autonomous systems. Valthos's investor syndicate (OpenAI, Founders Fund, Lux Capital) is equally unusual: AI companies and deep-tech venture firms are treating engineered biosecurity as a critical infrastructure layer rather than a niche defence vertical.

Corporate R&D and open-source: Varro Life Sciences' decision to open-source its core biosensor technology is the clearest signal for investors. An open-source sensor platform means any hardware integrator, from Boston Dynamics to a garage workshop, can build a compatible detector. This accelerates ecosystem development at the cost of direct product moat. The bet is that adoption volume will outrun the competitive advantage of proprietary lock-in, a strategy that worked for Android in mobile and is unproven in biosecurity.

The limits: what still needs to be solved

Why it works anyway: the dual-use market pull

The barriers above are real, but they are being addressed by a structural push that the technology alone does not need to solve: the market for autonomous threat detection is growing from both the defence and civilian sides simultaneously, and the overlap between the two is where the unit economics improve.

On the defence side, the U.S. Army's CSIRP transition to a program of record means that at least one procurement pipeline is locked in. The NATO Innovation Fund's Portal Biotech investment opens a second, transatlantic channel. These are not speculative grants. They are multiyear acquisition programs with defined delivery milestones and budget lines.

On the civilian side, the same sensors that detect anthrax at a military checkpoint can detect SARS-CoV-2 in a hospital lobby or industrial chemicals in a factory air handling system. Varro's open-source strategy is designed specifically to capture both markets with the same core technology, differentiating only in the software that interprets the data. The unified sensor market is larger than the defence-only market by a factor of 5–10×, which means startups that can satisfy both use cases have a path to scale that pure defence contractors lack.

The convergence is still early. No company in this space has more than 50 employees, and none has achieved FDA clearance or a production-rate defence contract. Between the protein engineering maturity, the platform availability, and the funding from both NATO and venture capital, the 2025–2027 window is when autonomous CBRN biosensing becomes an investable category rather than a research curiosity.