Automation Xperience takes place at a time when robotisation in manufacturing is accelerating, while the requirements for machine safety are also changing. New robot applications, collaborative systems, AI functionality and industrial security are making the safety issue more complex. For Pilz Netherlands, this is an important theme at the trade fair. Rocking Robots spoke about this with Jelger Slim, Safety Consultant and trainer at Pilz Netherlands.
“There is work to be done,” says Slim. “The new Machinery Regulation has already been published and will become mandatory in 2027. At the same time, the new parts of safety standards EN ISO 10218-1 and EN ISO 10218-2 were released in 2025. Companies that design, integrate or use robots need to start preparing now.”
“A lot is happening in the development of standards,” he says. “The new EN ISO 10218-1 and EN ISO 10218-2 were released in 2025. They pay considerable attention to the interaction between robots and people. From that theoretical framework, there are clear developments and new safety requirements.”
However, practice does not always keep pace with standardisation. “I see major differences between manufacturers, integrators and end users. I still come across robot applications where not enough thought has been given to the safety concept, mechanically, electrically and in terms of control systems. There is still a lot of room for improvement.”
New EN ISO Standards
According to Slim, the new EN ISO 10218 series is important because it better reflects how robots are used today. “The world of robotics has changed. In the past, the focus was mainly on the traditional industrial robot. With the arrival of collaborative robot applications, interaction between humans and machines has become much more important. The new standard provides a clearer reference framework for different types of applications and the safety measures that belong to them.”
“The new robot classification distinguishes between class 1 and class 2 robots, with factors such as speed, the forces released during movements and the weight attached to the robot playing an important role,” he says. “Payload also matters. With a payload of ten kilograms or more, a robot application may fall into a different class.”
According to Slim, this classification is not merely an administrative step. It also helps determine which safety measures are required. “The division into classes is intended to give designers, manufacturers and other parties involved a more workable framework. Previously, the distinction was less clear. Now it is easier to see which measures fit which application.”
Safety Does Not Stop at Purchase
This is also relevant for end users, because robot safety does not end with the purchase of a robot. A robot may comply with the manufacturer’s requirements as a product, but once it is integrated into a wider production line or robot cell, the safety assessment changes.
“As an end user, you may buy a robot from a well-known manufacturer,” says Slim. “But once you integrate that robot into a larger assembly of machines, you need to look again at the environment, the application and the conditions in which the robot actually operates. In practice, steps are still often skipped at that point.”
According to Slim, this issue often emerges during commissioning. A machine or robot may be purchased with the correct documentation, but settings are later adjusted, processes are changed or operators push the limits on the shop floor. As a result, an application that was initially safe can still become risky.
Process
“Things regularly go wrong in the transition from procurement to implementation and commissioning,” says Slim. “In practice, I see robot applications where insufficient account has been taken of all the safety requirements that apply in the final end-user environment. The robot does not stand alone; it is part of a larger assembly of machines. That context has to be included.”
“Manufacturers, designers, importers, distributors and end users will all have to deal with the new requirements,” says Slim. “Every party must understand its own responsibility. The Machinery Regulation and the associated safety standards require a chain-based approach.”
Cybersecurity and AI
One topic becoming increasingly prominent is cybersecurity. Modern robot cells and production lines are usually connected to networks. This enables remote control, data analysis and integration with other systems, but it also introduces new risks. The new Machinery Regulation explicitly addresses digital risks and industrial security.
“In the past, a robot application often operated locally and in isolation,” says Slim. “Today, entire production lines can be accessed or modified remotely. That means you have to think about unauthorised access, cyberattacks and the risk of someone shutting down or taking over systems.”
According to Slim, cybersecurity should therefore become part of the risk assessment. “Security risks must be included in the risk assessment. This mainly concerns the question of who has access to the system, which rights that person has and how you prevent unauthorised parties from influencing the machine.”
Responsibility
This becomes even more important with the rise of AI in production environments. AI functionality can make machines more adaptive, but it also raises new questions about predictability, responsibility and conformity assessment. “Artificial intelligence was not included in the old Machinery Directive in this way,” says Slim. “The new Machinery Regulation does take it into account. If AI becomes part of a machine, that can have consequences for the conformity assessment.”
This may mean that a notified body, also known as a NOBO, has to be involved. “When a machine is equipped with AI, there is a greater chance that an external party will have to assess safety and compliance. A manufacturer or end user will then need to draw up additional documentation, define testing and validation methods and be able to demonstrate that the applicable requirements have been met.”
Here too, responsibility does not lie only with the original manufacturer. “If a manufacturer adds AI to a machine, it must be included in the design process. But if an end user modifies an existing machine by adding AI functionality, that end user may also take on new responsibilities.”
Robots
“With robot applications that are combined with artificial intelligence, you are therefore dealing with new safety requirements and conformity issues,” says Slim. “That is certainly the case when such a robot becomes part of a larger assembly of machines. You must be able to demonstrate that the application as a whole is safe and compliant.”
For companies that use robots or plan to do so, Slim has practical advice. “Familiarise yourself with the Machinery Regulation and identify which changes are relevant to your product, system or application,” he says. “Carry out a gap analysis. Look at which new requirements affect your situation and determine which standards you can use to demonstrate conformity.”
That analysis must go beyond a legal check. “You have to look at the entire safety concept. What mechanical risks are there? What electrical risks? What control-system risks? And what security risks arise because systems are connected or can be accessed remotely?”
One complicating factor is that the new practical guide for the Machinery Regulation is still being developed. The European Commission is working on a new guide for the application of Regulation (EU) 2023/1230. It is expected to be available by the end of 2026. Slim considers that a tight schedule.
Clarity
“The practical guide should help companies interpret the new requirements properly. But we are now in 2026 and the Machinery Regulation becomes mandatory in January 2027. For manufacturers, designers and end users, that is a narrow timeframe.”
He understands that the process is complex, but also sees a practical problem. “Many parties are involved in such a guide and the content has to be drafted carefully. At the same time, companies need to prepare now. If the guide only becomes available late in the process, it will feel too late for many parties.”
That is why, according to Slim, companies should not wait for full clarity. “2027 may still seem some way off, but it is just around the corner. Companies that combine robot applications with AI, cybersecurity issues or complex integration into production lines in particular really have work to do.”
Rocking Robots will once again report as a media partner this year. Watch last year’s report here:
