This tweet captures the current state of readiness of robots for outdoor security well. Seemingly a good idea on the surface, send an army of 'bots out to patrol for safety, and deliver food, the reality is that the self-driving and environmental navigation algorithms that make these devices function (or not) still have a lot of development to do before the concept of robots serving society and navigating public spaces becomes more than a very expenisve experiment.
Automated robots failing to deliver some late-night takeout is certainly a first world problem with minimal real downside. Automated robots operating under the guise of security that can be easily disabled by preventing their movement via everyday objects, whether placed accidentally or maliciously, is a much larger issue. This example of the delivery robots causing a traffic jam from a simple unexpected blockade is just as applicable to security robots, and potentially security drones as well. These devices need live human operators overseeing their movements and providing additional guidance, which eliminates many of the claimed benefits and cost reductions of switching to automated patrols. We are unlikely to see a robot or drone takeover of the physical security space in the near future, and it might be more likely we see the eventual solution come out of a BattelBots episode than a Silican Valley startup.
Indoor spaces are a slightly different issue. As Cobalt Robotics is showing, there is a much stronger value proposition for robots patrolling indoor spaces, where the overall environment is under better control and the robots do not need to deal with all of the randomness of public spaces. Also, the applications are different, operating inside of an already controlled space allows the robots purpose to be better defined and more targeted to viable applications instead of being a PTZ-on-wheels hoping to spot a crime in progress.