AirNode now operates as SOVAIR.AI, a Strong Interactive venture in Miami Beach that builds air-gapped AI for regulated industries. A utility pilot is still in development.
Air-gapped AI for regulated industries has a new name, and it comes from Miami Beach.
AirNode, a venture of Strong Interactive, now operates as SOVAIR.AI. The company says the product, the team and the security model are unchanged.
Only the name has changed, according to the company’s release. MAIN asked founder Jonathan Herman for the background.
Why Did AirNode Change Its Name?
Herman told MAIN why. “When we ran a trademark clearance on AirNode we found three live third-party AIRNODE registrations, so that was a confusion we didn’t need.”
“We also wanted a name that better describes what we do, and SOVAIR comes from the two things our product is: sovereign and air-gapped,” he said.
What Is Air-Gapped AI?
SOVAIR.AI describes it as sovereign AI that is physically isolated. Models, compute and data stay inside the customer’s own security boundary.
The release says organizations with the most to gain from modern AI are often the ones barred from using it.
The company says utilities, hospitals, financial institutions and government agencies often cannot send sensitive data to a cloud AI service.
SOVAIR.AI does not try to secure the path to the cloud. It removes the cloud entirely.
What SOVAIR.AI Builds
Its Full-Stack Solution runs model inference, compute and document retrieval on-premise.
It includes a secure workbench with role-based access control and full audit logging.
Updates arrive as cryptographically signed offline manifests, so the system never opens a network path.
“Data stays inside the customer’s boundary, the models run locally, and nothing we build touches plant control,” said Jonathan Herman, co-founder and CEO.
The Utility Pilot
The first pilot is with a Southeastern U.S. utility, which the company does not name. It is a live pilot that is still being developed, Herman said.
The pilot is helping define the final product, the release says.
“The pilot is shaping how our solution fits a real utility’s environment, with various legacy systems that have existed for decades,” Herman told MAIN.
Data access is limited by design. Herman said there is some data that can be accessed using the tool and other data that cannot, by design, due to the legal requirements for safeguarding it.
On timing, he said nuclear-sector timelines run in months rather than weeks, and more updates will follow.
The use cases are advisory and built around documents. They include compliance support, inspection-readiness preparation, document review, secure knowledge search and tabletop exercise support.
The platform does not control plant systems. It is not used in any safety-related capacity.
The release adds that the leadership team includes two decades of cybersecurity management at one of America’s largest nuclear operators.
The Threat Weekly
Since August 14, SOVAIR.AI has published SOVAIR AI Threat Weekly, an executive briefing on AI-enabled cyber threats to nuclear, energy and other regulated sectors.
Herman says eight editions have run. The briefing is compiled from open-source reporting and is available on the company’s site and by RSS.
Asked who reads it, Herman gave an honest answer: “it’s early, and we’re still building the readership.”
“It’s written for the people who own the security boundary at a plant or utility, meaning CISOs, OT security leads and compliance teams.”
The company also publishes AI Jockey, a broader briefing. Herman says AI Jockey has about 2,200 readers across the Strong Interactive site and LinkedIn.
Recent editions covered a frontier AI lab’s confirmation that a model reached external systems during an evaluation.
They also covered a joint federal advisory on AI-generated exploit scripts aimed at industrial controllers.
A third item cited survey data showing that only a small share of operational-technology network segments are fully isolated.
“Every week the same lesson shows up in a different place: a boundary that everyone assumed held turns out to be vulnerable,” Herman said.
Where It Sits in Miami’s Security Field
Strong Interactive is based in Miami Beach. MAIN asked about its move from blockchain to AI.
Herman said Strong Interactive is “a studio with several ventures, and SOVAIR is the one built around our nuclear-cybersecurity co-founder.”
Miami has other security-minded AI companies. MAIN's Q3 2026 tally counted Comp AI, which raised a $34 million Series A for AI-powered compliance and security.
The same tally counted Cantina, which raised $8 million for an agentic security platform, with Framework Ventures investing.
MAIN has covered the field before in five Miami cybersecurity AI firms holding the line.
What Comes Next
Energy and utilities are the starting market. The company plans to expand into healthcare, finance and legal, where compliance determines what gets bought.
The pitch is isolation, not integration. The next proof is a pilot that reaches deployment.
Sources:
- SOVAIR.AI press release, October 5, 2026
- Jonathan Herman, email and written answers to MAIN, October 2026
- AI Jockey, Strong Interactive
- SOVAIR AI Threat Weekly
- MAIN: 5 Miami cybersecurity AI firms holding the line
- MAIN Q3 2026 Miami startup funding tally (Comp AI, Cantina)