Fleet outlook: Gecko sees per-unit invariants from OpenShell as a foundation for reasoning about swarm/fleet dynamics in future deployments.
- Platform launch: NVIDIA Corp. announced the Open Agent Safety Platform, comprising OpenShell and Sentry, to strengthen governance and control across hardware, compute, and software for AI-powered robots.
- Enforcement mechanics: OpenShell provides a secure runtime boundary that traces all actions and enforces policy, while Sentry can quarantine out-of-bounds agents in milliseconds; a Policy Proover validates the master decision tree against unintended exfiltration.
- Adoption scale: Gecko Robotics Inc. is among more than 100 organizations working with NVIDIA's technology, applying it to inspection robots that climb, crawl, fly, and swim on critical infrastructure.
- Hardware separation: Gecko's Komodo robot places an independent enforcement layer between the AI agent and the hardware, keeping agents within human-defined boundaries.
Gecko Robotics Inc. is using NVIDIA's new agentic AI security layer in its inspection robots, becoming one of the more than 100 organizations working with NVIDIA Corp. on the technology. For manufacturers watching AI autonomy creep onto physical equipment, this is one of the first concrete examples of enforceable, infrastructure-level control over what an AI agent can actually do to a machine.
A Full-Stack Safety Play
NVIDIA announced its Open Agent Safety Platform, which includes OpenShell and Sentry. The platform is designed to strengthen governance and control across hardware, compute, and software for AI-powered robots.
OpenShell provides a secure runtime boundary that traces all actions and enforces policy for AI agents. Because it is open-source, OpenShell can be extended to work with third-party compute platforms, including those from Arm and Intel — a meaningful detail for any plant running heterogeneous silicon. NVIDIA Sentry monitors agent behavior and can quarantine agents that attempt to move outside their boundaries in milliseconds. A Policy Proover in the platform validates the master decision tree to find unintended exfiltration.
"As we continue to discover the frontier of AI capabilities, we must accelerate discovery at the frontier of AI safety," said Jensen Huang, CEO of NVIDIA.
"For the agent economy to grow, we need a trusted foundation across silicon and software," said Justin Boitano, vice president of enterprise AI at NVIDIA.
From Software Guardrails to Physical Machines
Gecko Robotics uses robots that can climb, crawl, fly, and swim on critical infrastructure. That makes the control question more than academic: when an agent can command motion, a policy violation is not a bad API call — it is a physical event.
The developer defines intent; the enforcement layer holds the line.
"The idea that losing control of AI is inevitable is a dangerous excuse for inaction. We have a responsibility to build safeguards that keep AI within the boundaries humans set," said Jake Loosararian, co-founder and CEO of Gecko Robotics.
That framing — safety as an engineering discipline, not an inevitability debate — is the right posture for industrial deployment.
From One Robot to the Fleet
Gecko's Cantilever AI-powered platform transforms datasets into actionable insights, and the company's systems enable data-driven, evidence-based decisions on structures. OpenShell sits within the field-system boundary, governing agent behavior at the machine level.
The longer-term opportunity is scale. Deterministic per-unit guarantees are a prerequisite for reasoning about coordinated systems.
"It is certainly helpful to have invariants at the individual unit level that can be used to reason about swarm/fleet dynamics," said Ariel Weingarten, director of engineering at Gecko.
The takeaway for manufacturing professionals: agentic AI on physical assets is moving from model-level promises to infrastructure-enforced policy. That is the difference between trusting a system and verifying one.
Eugene Demaitre is editorial director of the robotics group at Arrowfly.
Is this your company?
This article features your business. Claim it to add your logo, contact details, and a link to your website — or upgrade to reach more buyers.
Did you know 80% of Press Releases trigger AI content warnings? Reach out and the M4S team can assist.