The traditional industrial robot lives in a cage. It is fast, strong and utterly indifferent to what is in its path, so the safety model is separation: put a fence around it, interlock the gate, and keep people out while it moves. Collaborative robots — cobots — were built to escape the cage and work in the same space as people, and that single change moves the entire engineering problem from how well the robot performs its task to how reliably it does the right thing when something goes wrong. A cobot's defining feature is not its dexterity. It is that it can be trusted to stop.
That trust is not a marketing claim; it is a set of certified safety functions with standards behind them. Force and power limiting so that contact with a person stays below a threshold that causes injury. Speed and separation monitoring so the robot slows as a person approaches and halts if they get too close. Safe, monitored stopping that is proven to work even when the main controller misbehaves. These functions are what a safety assessor examines, and they are the difference between a robot that is allowed to share a workspace and one that is only allowed behind a fence with a faster cousin.
The point that trips up first-time adopters is that safety is a property of the application, not of the robot. The same arm is safe or unsafe depending on what tool it holds, how fast it moves, what it carries and how close people stand — a robot handling a blunt, light part at low speed can be genuinely safe alongside a worker, while the identical robot holding a blade or moving a heavy load cannot. So every deployment requires its own risk assessment, and a certificate on the robot is not a certificate on the installation. This is exactly the expertise that a small manufacturer, the natural customer for a cobot, is least likely to have in-house.
That gap is shaping how the technology reaches the market. Rather than selling a bare arm and leaving the safety analysis to a buyer without a safety engineer, vendors and integrators increasingly deliver an assessed, application-specific safety envelope: the robot, the tooling, the speeds and the layout evaluated together, with the documentation that a regulator or an insurer will ask for. The robotics story here is not about a machine getting cleverer. It is about the unglamorous discipline of proving, in advance and on paper, that a machine sharing a room with a person will fail safely — because that proof, not the machine's capability, is what determines whether it can be there at all.