FORT Robotics adds “outside-in” safety features as it joins NVIDIA Halos for Robotics ecosystem

FORT Robotics, a company focused on safety for autonomous robots, said it is extending its “trust layer for physical AI” by adding an “outside-in” safety capability and joining the NVIDIA Halos for Robotics ecosystem. NVIDIA’s Halos for Robotics is positioned as a way to bring safety and reliability tools to robots operating in dynamic, physical environments, and FORT said it will demonstrate an “agentic safety” application built around that approach.

The announcement frames FORT’s work around “physical AI,” meaning AI systems intended to operate in the real world rather than only in digital settings. In that context, FORT characterizes its product as a trust layer designed to improve how robotic agents behave when facing uncertainty, potentially reducing the gap between what an AI model can do in testing and how it performs when it is embedded in day-to-day operations.

FORT said its new contribution is “Outside-In Safety.” While the company did not provide additional technical details in the announcement available here, the phrase suggests a safety method that looks at external conditions or environmental indicates to constrain or guide the behavior of an AI-driven robot. FORT’s stated goal is to apply that concept as part of an agentic safety application, according to the announcement.

NVIDIA Halos for Robotics is described in the announcement as an ecosystem that brings safety to autonomous robots. By joining it, FORT is aligning its safety layer with NVIDIA’s broader robotics initiative, which is intended to support developers working on robotics deployments that require more than just autonomy. The partnership also indicates that safety tooling, not only robot autonomy, is becoming a core focus in robotics ecosystems.

The practical significance for robotics developers is that “trust” and “safety” are increasingly treated as components that must integrate with AI decision-making. As robots move from controlled lab settings toward warehouses, manufacturing floors, and other environments with people and physical constraints, vendors and ecosystem builders are under pressure to address safety during real-time operation, not only during pre-deployment validation.

Even with the collaboration, the specifics of how FORT’s outside-in safety works, what sensors or safety indicates it uses, and what performance metrics it improves were not disclosed in the available announcement text. The company also did not provide timeline details on availability, customer pilots, or measurable outcomes from prior deployments, limiting what can be confirmed about scope and readiness.

For NVIDIA, this kind of ecosystem expansion fits a broader pattern in the industry: big AI platforms increasingly rely on third-party safety and robotics tooling to support real-world deployments. For startups like FORT, aligning with a major AI vendor’s robotics program may help with integrations and credibility, even when the market still has unanswered questions about standards, certification, and how different safety approaches compare in practice.

The next item to watch is whether FORT’s “agentic safety” demonstration includes concrete details such as failure modes the system is designed to prevent, the operational boundaries it enforces, and any benchmarking methodology. Another key question is how the outside-in safety layer integrates with NVIDIA’s robotics ecosystem in a way that developers can reproduce across robot platforms and deployment scenarios.