Humanoid Robots Enter Factories, and Liability Lands on HR

A robot may pass a performance test and still become a safety hazard the moment it meets an unexpected obstacle, a software update or a real worker. That gap between test success and workplace safety is now the central problem for employers deploying humanoid robots on factory floors, according to industry figures who say accountability must be shared among manufacturers, developers, integrators and the companies using the machines.

The warning carries practical weight for HR and workplace safety leaders. A single robot error can pull four parties into a liability dispute, and employers sit squarely among them, responsible for training workers, keeping safeguards in place and maintaining procedures to stop unsafe operations.

What triggered the debate?

Agility Robotics and FORT Robotics announced a partnership in October to strengthen the safety architecture of Agility's Digit 5 humanoid robot. Their approach includes onboard safety systems, external interfaces, communications and manual override, according to the source.

The development signals a shift, the piece argues. Robots must now be safe in human workplaces, and that requirement extends beyond the machine itself to the systems and people around it.

Technology alone cannot settle the accountability question, however. When a robot acts on a faulty component misidentified by an AI system, the manufacturer, the software developer, the system integrator and the employer could all face liability.

Why do humanoids create new physical risks?

Traditional industrial robots work inside enclosed cells, isolated from workers by barriers and safety devices. Humanoid robots do the opposite. They navigate environments built for humans, perform varied tasks and adapt to changing conditions.

That flexibility produces new hazards, the analysis states:

• A false detection can trigger an unnecessary physical movement.
• A software update can alter a robot's behavior.
• Changes in factory layout can create hazards absent from earlier tests.

Dana R. Wolfe, Vice President and Head of Industry at GFT Americas, told Unite AI that preventing unsafe interventions requires multiple layers of protection, including deterministic software guardrails, factory-specific confidence thresholds, sensor fusion and staged validation before deployment.

"There is no way to completely avoid false-positive results," Wolfe said. Instead, manufacturers must ensure that mistakes do not translate into unsafe physical movements. Confidence thresholds, human authority and hardware interlocks should all sit inside the system's safety architecture, she said.

Do current standards cover these machines?

Not fully, according to the analysis. The International Organization for Standardization has published safety requirements for industrial robots, and ISO 10218-2 covers the integration of robots into complete systems. But these standards do not address all applications or hazards of mobile robots working alongside humans.

Independent testing, scenario-based validation and phased workplace pilots matter most as industry practices evolve, Wolfe said. Safety evaluations must also be repeated whenever changes to the AI model, software, movement restrictions, tasks or factory environment could invalidate earlier results. A robot that was safe under one set of conditions may not remain safe after a major change.

What should human oversight actually mean?

Adrian Figueroa Ramos, co-founder and CEO of Flexio and Routina, argued that oversight should target the decisions carrying the most uncertainty and risk.

"In today's AI world, the volume of outputs is more than a human can process. Reviewing everything makes the human the bottleneck, but human judgment is still necessary for high-risk or low-confidence decisions," he told Unite AI.

For physical AI, that principle translates into concrete HR obligations. Human oversight should involve named individuals with the power to override a system, and decisions should be documented and checked rather than rubber-stamped through a single approval click.

Who pays when a robot fails?

The analysis proposes distributing accountability according to each party's control over the system:

• Manufacturers create and document safety measures.
• Software developers test changes that may affect perception or movement.
• Integrators validate the whole system in its intended environment.
• Employers train workers, ensure safeguards and establish procedures for stopping unsafe operations.

Cybersecurity compounds the exposure. Operational technology tied to robots can be vulnerable through sensors, software and external systems, and a security failure that affects physical behavior is a workplace safety problem, not merely an IT problem, the analysis notes.

Humanoid robots are expected to grow more flexible and productive in factories, but adoption will not rest on performance metrics alone. The source closes on a defining question: a robot can make mistakes, but only if the people using it have built a system that holds them accountable for those errors.