Summary Since 2010, industry has been using a new type of robot capable of interacting with workers during production or simply sharing the same workspace with them. They are called collaborative robots, or cobots. Cobots differ from conventional robots with respect to machine-worker interaction; they accompany workers, assist them and help them perform tasks. However, this technological change also brings with it new risks, especially in production: collision risks (since humans can come into contact with cobots), risks of musculoskeletal disorders (MSDs) (even though a cobot may be designed to prevent such problems, operators must learn to handle it properly to limit or prevent them), psychological and social risks (human stress related to cobot movements and work pace), etc. An exploratory study was conducted to investigate these issues. The purpose of the study was, first, to make occupational safety recommendations regarding the implementation of robots in a collaborative setting and, second, to suggest ways to inform stakeholders about the issues involved in implementing cobot installations. There were two parts to the study: a theoretical part focused on examining plans and a practical part focused on the actual situation in the field. The purpose of the first part was to assess how the robot’s safety functions, which are processed by safety-related electronic boards, ensure operator protection within the framework of the four modes of collaborative operation established by standard ISO 10218:2011: (1) safety-related monitored stop, (2) hand guiding, (3) speed and separation monitoring and (4) power and force limiting. According to standard ISO 13849-1:2015, a safety function of a machine is a function of the machine whose failure can result in an immediate increase in the risks to a person’s physical well-being. For example, the emergency stop on a machine is a safety function. The purpose of the second part of the study was to gather feedback on taking safety into account in collaborative robot integration projects in Quebec. Feedback was collected from three types of participants: collaborative robot users (employers), integrators and workers involved. The theoretical part was conducted by analysing the technical reference material for three robots: one designed to be collaborative and two conventional robots converted into cobots. In addition, a brief case study illustrated the implementation of a cobot installation based on the analysis. The field part of the study was carried out by observing cobot installations in four companies and conducting semistructured interviews with the participants concerned. The case study of the theoretical part showed that to make a mode of collaborative operation safe, a number of safety functions must be combined, depending on the robot. The theoretical part revealed, in particular, that the performance level set by the manufacturer for a safety-dedicated controller is not a guarantee of the overall performance level of the safety function processed by the controller. The controller’s electronic board only performs the “signal processing” part of the safety functions. An employer that procures a collaborative robot equipped with a controller that meets the performance level set in the standards should not assume that the cobot’s safety functions will necessarily meet the requirements of the standards. A robot’s safety functions often need to be supplemented by adding an input device, such as a presence-sensing device. In some cases, environmental constraints may be an overriding factor in the choice of a sensing device. Choosing the right device is crucial, because if its safety performances are below those of the processing part, the safety function will be less reliable. The requirements in the standards recommend so-called safety devices. The field part of the study revealed that cobotics is in its infancy in Quebec. Only one of the four companies visited had a cobot installation that was up and running. The other three were in the process of cobot integration. Other companies contacted by the study team when recruiting participants for the visits were either planning to procure a collaborative robot or were still thinking about how to integrate it into their production process. The companies visited chose their cobots on the basis of (1) their low cost compared with that of a conventional robot, (2) quick return on investment, (3) being able to reassign workers to more rewarding tasks, (4) space constraints and (5) potential occupational health and safety (OHS) risk reduction. Thus, to judge from what we observed, companies are turning to cobots not necessarily to meet a need for human-machine interaction in production, but rather for financial, spatial and OHS reasons. The integrators we met confirmed that designing a cobot installation is a complex task because the technology is very new and involves some very specific safety requirements. The most difficult step, they say, is the risk assessment. Often, it was limited to risk identification. But to know whether a robot can be used in a collaborative context, at the very least the risks of the future installation must be estimated in order to determine the minimum required performance level. The performance level of the safety functions must be equal or superior to this minimum and be consistent with the analysis of the risks associated with collaborative operation. Last, for safety to be a decisive factor when determining needs and integration, it must be included in each planned functionality of the installation. This inclusion must be implemented through a close dialogue involving the employer, the integrator and workers. From what we observed, workers were not very involved in the process of determining needs and integration. Given this low degree of worker participation in risk assessment, which in itself is a complex step, we are now proposing to conduct research on the cobot-related risk assessment process, by incorporating analysis of worker activity. In addition, regarding risk reduction, some aspects are worth investigating further, particularly (1) presence sensing, in modes 1 and 3, with non-safety devices and (2) the applicability and acceptability of the force limit values given in technical specification ISO/TS 15066:2016. Worker proximity to the robot during production requires a risk evaluation focusing on possible contact. In sum, using collaborative robots does not systematically mean that protective measures are no longer necessary. Risk assessment is always needed at the integration stage, as manufacturers’ reference manuals and cobotics standards specify. Depending on the degree of acceptability of the risk in question, protective measures may well be required.