Machine safety law which includes robots, cobots etc. used to mean one thing: keep bodies out of the way of moving parts. Guards, emergency stops, crush zone clearances, lockout procedures. If you asked a compliance officer what "machine safety" covered, mental health would not have made the list.
That's no longer accurate, and the shift happened fast enough that most people leading robot or AI rollouts haven't caught up.
Look at Australia. In February 2026, New South Wales passed the first Australian law to explicitly regulate AI and digital work systems as a workplace health and safety hazard. It doesn't only cover robots on a factory floor. It covers the algorithm that assigns your shifts, monitors your output, or decides who gets flagged for underperformance. Under the new rule, an employer is responsible for managing the psychosocial risk that system creates, excessive workload the algorithm sets, undue pressure from performance metrics, biased decisions, even when the software was bought from someone else entirely. You cannot point at the vendor. The obligation sits with whoever put the system to work on your people.
Europe moved just as fast. The EU's new Machinery Regulation, which fully applies from January 2027, rewrote the ergonomics requirement that machinery designers have to meet. The old rule asked designers to reduce physical strain on operators. The new one adds mental strain explicitly, discomfort, fatigue, and mental over-strain all have to be minimised, and the human machine interface has to be adapted to what a person can reasonably process, not just what they can physically reach or lift. The regulation leans on an existing engineering standard, ISO 10075-2, built to help designers avoid monotony and overload, and now brings in what occupational psychologists call technostress, the specific strain of adapting to a system that keeps changing on you. Regulators wrote this in specifically because of machines with self evolving, AI driven behaviour, cobots that adapt in ways a fixed industrial arm never did. The law now treats that unpredictability as a hazard to a person's mental state, not just their body.
Put those two next to each other and the pattern is hard to miss. One law says the machine itself has to be designed with your mental load in mind. The other says the company deploying it is legally accountable for the mental load that results. Neither treats psychological safety as a values statement or a culture initiative anymore. Both treat it as something a regulator can inspect.
This will reframe how you think about the phrase "machine safety." Lots of leaders still picture it as a fixed category, something engineering or facilities handles, cages and stop buttons and a checklist that ends at the physical world. That checklist just got longer, and it now includes the same question HR has been asking about burnout and monitoring for years, except now it comes with a legal enforcement mechanism attached to the hardware and software itself.
If you're introducing a robot, a cobot, or an AI system that assigns work, monitors performance, or interacts directly with your people, the mental load question isn't a soft add-on to the safety plan anymore. In at least two jurisdictions, it's now part of the actual legal definition of a safe machine.
Regulation (EU) 2023/1230 of the European Parliament and of the Council of 14 June 2023 on machinery, Annex III, Essential Health and Safety Requirement 1.1.6 (Ergonomics), applicable from 20 January 2027. Work Health and Safety Amendment (Digital Work Systems) Act 2026 (NSW), passed February 2026. ISO 10075-2:1996, Ergonomic principles related to mental workload, Part 2: Design principles.