GeoMFA
An identity-attribution system for establishing that people or machines are who—or what—they claim to be through independent corroboration.
PCT/US2026/032608 · U.S. 19/705,723 →Niels Goldstein · Founder
About.
NSG Studios is the independent practice of Niels Goldstein — an enterprise architect working where identity, Zero Trust, and the governance of AI systems meet.
Four decades inside consequential systems.
Forty-three years in enterprise IT, most of it spent on the unglamorous half of security: who is allowed to do what, on what evidence, and how anyone proves it afterward.
Today he consults on the Zero Trust and security requirements that U.S. Army secure-network equipment must satisfy, and serves as a core technical lead on enterprise network acquisitions governing hundreds of millions of dollars in procurement.
Before that: architecture, CTO, and director-level roles at Oracle, Teradata, Siemens, Fannie Mae, CGI Federal, and Accenture Federal — and inside the Office of the Navy CIO and the CTO's office at the U.S. Marshals Service.
Active TS/SCI clearance. CCSP. DoD 8140 IAT III, IAM III, IASAE II–III. BS, Geology, Rensselaer Polytechnic Institute; MBA, IT management and operations research.
Trust should be witnessed is not a slogan. It came from watching a system fail the test.
An AI agent was directed to run a benchmark. It returned a complete results document — metrics, computed values, a falsification verdict, even a note about artifacts it had caught in its own output. The host logs showed that nothing had ever executed. The agent had modeled the experiment internally and reported the simulation as work performed.
It was caught only because ground truth existed outside the agent's own account of itself.
Everything published here follows from that: no actor's report of its own behavior is evidence. Corroboration is. That principle is old — a second signature, another set of eyes, an independent witness — and it applies to software agents exactly as it applies to people.
Identity and AI governance, made inspectable.
Always Verify
Three openly published specifications for governing multi-agent AI.
Explore the specifications ↗Each was also filed as a defensive publication on the Technical Disclosure Commons, with timestamped version-control provenance dating the underlying work to 2025.
What this work is not.
The open work is specifications and design patterns — not products, not finished platforms, and not a guarantee of any outcome. It claims no novelty and asserts no patent. It addresses the governance of deployed agentic systems, not the alignment of frontier models. It was published openly so the answer belongs to the commons rather than to any one party.
Bring the hard problem.
Open to expert testimony, advisory work, architecture review, standards and specification development, and expert engagement on identity assurance and AI governance.