2006 vs 2026 — everything changed except the safety layer

In 2006, the typical industrial robot had a PLC for a brain. It executed pre-programmed motion sequences inside a physical cage. No humans were supposed to be nearby during operation. The Machinery Directive 2006/42/EC defined the safety requirements, and the safety layer was simple: hardware interlocks, emergency stop circuits, light curtains, physical barriers.

Twenty years later, the picture looks different in every dimension except one.

The brain changed. Many new robotic systems use neural networks for perception, planning, or control — often alongside classical control algorithms. The robot doesn't just follow a pre-programmed path anymore. It adapts. It responds to sensor input in real time. In some cases, its behaviour evolves through machine learning.

The workspace changed. Collaborative robots now share space with human workers. The cage is gone or reduced. The distance between the robot's actuator and a human body dropped from meters to centimeters. The IFR reports that collaborative robot installations have grown every year since 2017.

The regulation changed. The EU replaced the 2006 directive with Regulation 2023/1230, adding requirements for cybersecurity, AI-based safety functions, autonomous mobile machinery, and post-deployment software updates. The regulation applies from 20 January 2027 — with no transition period.

The safety layer didn't change. The fundamental architecture — hardware interlocks, emergency stops, safety-rated PLCs — remains largely the same technology that was designed for deterministic, caged systems. These mechanisms work well for what they were built for. They were not built for systems where the AI might make a decision the programmer didn't anticipate.

That's the gap. The brain changed, the workspace changed, the regulation changed. The safety layer underneath is still designed for the world of 2006.