Robots Are Not 100% Safe, and Employers Must Plan for That

Industrial robots will never be 100% safe, and the risks they carry are shifting onto employers rather than disappearing. That is the central argument of Stephen J. Bigelow, senior technology editor at TechTarget, who has spent more than 30 years writing about technology and who now warns HR and business leaders against treating embodied AI as inherently safer than human workers.

His case rests on a reporting gap. Mandatory public reporting of robot accidents is largely absent, and incident insights are rarely compiled into usable statistics, Bigelow writes. As a result, no reliable count exists of warehouse robots dropping pallets, surgical robots injuring patients, or other embodied AI incidents. One number does exist: the National Highway Traffic Safety Administration tracked roughly 825 automated driving system incidents nationally by April 2026.

What does the anecdotal record show?

Even without statistics, humanoid robots drew significant bad press in 2026. Academic research compiled a list of embodied AI mishaps:

• A remotely operated humanoid frightened an elderly woman in Macau.
• A humanoid embraced a dancer during a university performance in Xi'an.
• A robot allegedly kicked a boy during a demonstration in Urumqi.
• A humanoid caught fire after crashing into a barrier at the World Humanoid Robot Games in Beijing.
• German police seized a humanoid in Cologne after it allegedly touched a 16-year-old girl.

News footage has also shown Waymo vehicles blocking traffic on busy city streets.

Bigelow does not dispute the value of robots at work. In hazardous environments, they take over dirty, dangerous and demeaning jobs, shielding workers from extreme heat, toxic chemicals, heavy molten metal and radiation. Machines carrying heavy loads over long distances reduce severe traumatic injuries such as broken bones and falls. Automating repetitive twisting, hammering and manual assembly lowers the risk of musculoskeletal disorders, including carpal tunnel syndrome, back pain and joint wear.

Why does the risk remain?

The danger, he argues, lies in the machinery itself and its digital exposure. Robots are complex, extremely heavy and mechanically unstable, raising the risk of malfunctions that can cause severe injury or death. Machine learning software and network connectivity open the door to software errors and hacking. Unpredictable real-world environments add further uncertainty. "Uncertainty is risk, and the idea of robot safety is simply a euphemism for shifting risks as robots become realities in our workplaces," Bigelow writes.

Progress is real. Fenceless guarding uses sensors to protect humans. Fall mitigation forces a robot to stop or sit when a person approaches. Cobots use force-limiting joints and elastic actuation to soften collisions. Wireless networking relies on isolated, encrypted links. The emerging ISO 25785-1 standard frames the design of dynamically stable humanoid robots.

What will safety improvements cost?

Bigelow predicts robot safety will follow the trajectory of automotive safety, where every crash gets analyzed and new materials, designs and standards evolve in response. The cost, he cautions, is human suffering. Airbags, crumple zones and anti-lock brakes followed decades of deaths. Two-hand start buttons, emergency stops and lock-out tag-out came only after industrial workers lost lives and livelihoods.

"Every laser fence, every camera, every emergency stop or kill switch, every behavioral model update, every line of reflective tape along a factory floor owes its provenance to human injury," he writes. Robots, like cars, are not safer; the consequences of their accidents are mitigated.

Can reporting accelerate progress?

Secrecy slows improvement, Bigelow argues. Businesses that use robots stay quiet when accidents occur. OSHA may receive a workplace report, or NHTSA may learn of an autonomous vehicle accident, but there is little incentive to share circumstances and lessons learned. He points to Safety Ledger as an early example of a centralized platform documenting injuries, deaths and property damage tied to autonomous systems and AI technology.

Such reporting mechanisms are not universally mandated or centralized today. Bigelow's closing appeal is directed at exactly the executives weighing embodied AI in the workforce: robots are here now, and open, honest, detailed safety collaboration may be the key to making them safer sooner, protecting the humans they were designed to help.