Intuitive offers a training pathway which incorporates a course through their company and subsequent proctored cases at the physician's facility. Proctoring is performed by a surgeon already certified on the system.
Actual credentialing (read permission to perform robotic surgeries) is performed by the facility/hospital where the surgeon works. Most hospitals have a requirement that a surgeon perform a certain number of proctored cases with the robotic system before being allowed to operate independently. This number is not high; at one of the large hospitals where I work, a practicing surgeon only needs to perform 3 cases with the robot before being allowed to operate independently (as best I can remember). At another hospital in the program where I train, surgeons were required to perform 25+ laparoscopic cholecystectomies before being able to perform them independently when laparoscopic surgery first became a big deal. I believe at least one of the lawsuits against Intuitive (Taylor vs Intuitive) is based in part on the fact that Intuitive reps may be pushing hospitals to lower their credentialing requirements.
As a related topic, credentialing surgeons is a difficult issue. The topic of evaluating surgical proficiency is extremely important right now in the world of academic surgical education. Currently, a board certified general surgeon has met three (main) requirements: 1) Completing an accredited general surgery residency, 2) passing a qualifying exam (written exam on surgical knowledge), 3) passing a certifying exam (oral boards, a test of a surgeons clinical reasoning and knowledge). None of these requirements, with possibly the exception of the first, provide a true criterion based measure of a surgeons proficiency at the actual task of operating. There is a push in the world of academic surgery to develop methods for assessing how well surgeons actually operate and whether or not they are competent.
In an ideal world, surgeons would not operate independently with the da Vinci or any other surgical tool without proving some level of proficiency first, which should be based on ability not number of cases performed. Becoming proficient, however, requires operating when not proficient. The best practices for training in a manner that maximizes physician education and minimizes risk to patients are still undetermined and the area is open for more research and innovation.