· MagenTrust Research
MagenTrust brings continuous human verification to physical AI. With Cacti Industries' universal human interface and Verge's SentientDB via Carahsoft, verified command authority reaches robots, fleets, and critical infrastructure.
Continuous human verification is leaving the keyboard and entering the field.
MagenTrust was built on a simple observation: a credential proves who you were at login, but Zero Trust demands knowing who is acting right now. For two years we have applied that principle to sessions, scoring behavioral entropy continuously and returning a deterministic PASS, CHALLENGE, or INTERRUPT before a consequential action executes.
Today, the consequential actions are changing. They are not just wire transfers and admin commands. They are a robot arm receiving a voice instruction, an autonomous fleet re-tasked mid-shift, or a drone released into a corridor. AI has become physical, and the question of who authorized an action carries more weight than ever.
Physical AI needs command authority that can be verified at the moment of action, not assumed from a credential presented at the start of a shift.
Our partnership with Cacti Industries brings continuous identity assurance to the place where humans and physical AI meet. Cacti describes its platform as the Human Intent Layer for Physical AI: one software, any robot, any wearable. Its wearable-first software lets operators command and supervise mixed, vendor-agnostic fleets with voice and pointing gestures, including beyond-line-of-sight operations.
That capability spans mission planning, field fleet supervision, and natural command and control. It also creates a Zero Trust question at every level. If an operator can speak or point a command to a fleet, how does the system confirm that an authorized human is still present forty minutes into a shift?
This is the problem MagenTrust was designed to address. Timing, cadence, hesitation, and motion-derived interaction signals can be evaluated continuously and ephemerally, without building a biometric database or retaining PII. Every fleet command can carry a current human-confidence verdict. PASS allows the policy-controlled action to proceed. CHALLENGE requests a low-friction confirmation. INTERRUPT stops the action before execution.
Physical AI cannot rely on an honor system. Authority must be checked where intent becomes machine action, using policy the operator controls.
Verified command is only half of governance. The other half is durable evidence that remains available wherever the mission runs.
Our integration with Verge Technologies connects MagenTrust to SentientDB, Verge's AI-driven, self-managing database platform for hybrid and multicloud environments. For critical infrastructure operators, verification verdicts and authorization trails can remain available across on-premises, edge, and cloud deployments when auditors, incident responders, or insurers need to determine who authorized what.
SentientDB is available to public-sector buyers through Carahsoft on contract vehicles including NASA SEWP V, ITES-SW2, NASPO ValuePoint, NCPA, and OMNIA Partners. MagenTrust reaches those buyers through its SentientDB integration rather than through a direct MagenTrust contract vehicle.
Cacti works across defense reconnaissance and resupply, construction and heavy industry, agriculture, and mining. These are environments where physical AI governance is not hypothetical. Once autonomous systems move goods, operate equipment, and coordinate with human crews, errors may not be reversible and responsibility cannot be ambiguous.
The MagenTrust, Cacti, and Verge stack divides that problem into three enforceable layers:
Where autonomy should run without human interference, the model inverts. A Human Exclusion Zone declares a perimeter of verified machine-only execution where human behavioral signal triggers INTERRUPT. The same control plane supports governance in both directions.
There is a deeper reason this architecture belongs in the field. Behavioral entropy is event-driven by nature. A keystroke, a pointing gesture, a shift in gait: these are sparse, asynchronous events, not continuous data streams. That is precisely the workload neuromorphic computing was built for. Spiking neural architectures process events as they occur and sit idle between them, which means continuous human verification can run at milliwatt power budgets, on the wearable or the gateway itself.
For edge deployments, the implications compound. Verification verdicts render locally, with no cloud round trip, so PASS, CHALLENGE, and INTERRUPT hold up in disconnected, intermittent, and comms-denied environments, exactly where defense and critical infrastructure operations live. The trust decision happens at the point of action, and SentientDB synchronizes the evidence trail when connectivity returns.
We see neuromorphic edge processing as a key direction for making continuous assurance truly continuous: always on, always local, and indifferent to the network.
Physical AI does not wait for quarterly access reviews. It acts in seconds, in the world, alongside people. Organizations deploying it need identity assurance that operates at the same tempo: continuous, deterministic, and examinable.
That is what getting physical means to us: the same PASS, CHALLENGE, and INTERRUPT discipline that protects sessions, now protecting the moment a human tells a machine to act. MagenTrust deploys inside the customer perimeter, with no data egress, no PII, and no biometrics required or stored.