Why does a smarter robot need safety systems outside the robot?

Agility Robotics and FORT Robotics signed an October 1 memorandum of understanding to expand the safety infrastructure around Digit 5.

The companies describe a three-part architecture: Safety Pendant → On-robot Communications → Offboard Safety Interface. The offboard layer connects Digit to external safety systems in the facility rather than relying only on sensors inside the robot.

As humanoids move deeper into human workspaces, safety becomes harder to contain inside a single machine.

Traditional automation could rely heavily on separation

A strong industrial-safety pattern has been to place robots inside defined workcells and physically separate people from the machine.

Agility positions Digit 5 around cooperative safety and close-proximity work, with human detection and functional safety behaviors intended to reduce reliance on external physical barriers.

As the robot's freedom of movement increases, the area that the safety system must understand and manage grows as well.

Safety expands from Robot Feature to Facility Architecture

Real facilities contain people, other robots, conveyors, equipment, emergency stops and existing safety systems.

The Agility-FORT Offboard Safety Bridge is designed to connect environmental safety signals with the robot's onboard architecture, extending cooperative safety beyond the robot itself.

Banseog reads this as Onboard Safety → Offboard Safety → Facility Safety Architecture.

More autonomy can increase the need for independent manual intervention

Digit can operate autonomously during normal work, while FORT's Digit 5 pendant architecture combines emergency-stop and enabling-device functionality.

During setup, maintenance or unexpected situations, the pendant gives customers a redundant manual override and an independent layer of protection.

Industrial autonomy can therefore follow Autonomy ↑ → Operating Envelope ↑ → Safety Complexity ↑ → Independent Override rather than simply Autonomy ↑ → Human Control ↓.

65,000 hours and 100,000 totes turn safety into an operational problem

Agility says previous Digit generations have logged more than 65,000 operating hours across customer environments including GXO, Schaeffler, Amazon and Toyota Motor Manufacturing Canada.

At GXO's Georgia facility, Digit passed 100,000 cumulative totes and Agility reports approximately 98% accuracy while on task.

Commercial deployment turns safety from a demonstration question into an Operational Safety problem that must survive thousands of repeated interactions and changing environments.

As commercialization scales, safety becomes a deployment condition

Agility says that as of May 2026 it had more than $300 million in multi-year Digit 5 customer orders.

Those orders are subject to satisfaction of certain contractual milestones and should not be read as $300 million of already recognized revenue.

As buyers move from pilots toward fleet deployment, the question shifts from whether the robot can perform a task to whether the entire facility can operate that robot safely at scale.

The buying unit can grow from Robot to Robot + Safety Architecture

Payload, runtime and walking speed are not sufficient procurement criteria for a humanoid operating near people.

Buyers also need to understand human detection, facility emergency-stop integration, external safety signals, safe state under communications failure and independent intervention during maintenance.

The buying unit can therefore expand from Robot to Robot + Safety Architecture + Facility Integration.

Safety can expand the operating envelope rather than only restrict the robot

Safety is often framed as a mechanism for slowing or stopping machines.

But well-designed cooperative safety can allow a robot to operate in a broader set of spaces and workflows closer to people. Agility and FORT explicitly connect onboard safety with offboard communication to extend the safety envelope.

Safety can therefore become a capability that expands the commercially usable operating envelope.

The talent stack extends beyond Robot AI Engineers

The partnership scope includes joint hardware development, solutions engineering, regulatory compliance and deployment support.

As humanoid deployment grows, Functional Safety, Systems Engineering, Industrial Integration, Regulatory/Compliance and Field Deployment can become increasingly important capabilities.

Industrial Physical AI requires not only intelligent models but people who can make those models operate repeatedly, safely and compliantly alongside existing systems.

BANSEOG VIEW | In Physical AI, Safety becomes Infrastructure

The Agility-FORT relationship has deepened from a custom component to a safety pendant, on-robot communication, offboard interface and strategic safety partnership.

Banseog reads the shift as Autonomy → Operating Envelope → Human Proximity → Safety Complexity → Facility Integration → Deployment Scale.

The smarter the humanoid becomes, the less sufficient robot intelligence alone is. The facility around the robot increasingly becomes part of the safety system required to deploy Physical AI at scale.

Banseog View — Autonomy → Human Proximity → Safety Complexity → Facility Integration → Deployment Scale

As humanoids enter human workspaces, safety can expand from onboard features into offboard and facility infrastructure.

Safety can become a deployment capability that expands the operating envelope rather than only restricting motion.

The Physical AI talent stack can extend toward Functional Safety, Systems Engineering, Industrial Integration, Compliance and Field Deployment.

Primary sources and references

The $300M+ figure is Agility's reported multi-year Digit 5 customer orders as of May 2026 and is subject to satisfaction of certain contractual milestones; it is not equivalent to already recognized revenue. The 65,000-hour, 100,000-tote and ~98% figures are company-reported operating metrics. Onboard Safety → Offboard Safety → Facility Safety Architecture and Autonomy → Operating Envelope → Human Proximity → Safety Complexity → Facility Integration → Deployment Scale are Banseog analytical frames.