Today, Boston Dynamics’ robotic dogs – which cost up to $300,000 each – already exist. Rotating data centers Across the United States, guarding the infrastructure that supports big tech companies’ generative AI, the companies building the world’s most powerful AI are entrusting its protection to robots. In reality, AI is guarded by robots that operate on AI themselves
There is an important detail in this arrangement: these robots do not make a single decision alone. Their role is strict surveillance: monitoring, patrolling and detecting anomalies. They do not implement force or independent action. Any response to the perceived threat remains in human hands.
Robots that do not make autonomous decisions reflect a conscious choice on the part of developers and operators around the world. This limitation is not the decision of a single company or jurisdiction. It’s a common principle that has emerged independently across regions – from Dubai to New York, across Chinese cities and US data centres. Wherever these systems are deployed, the boundaries remain the same: robots monitor, humans decide.
Infrastructure boom and new demand for surveillance robots
American technology companies are Investing hundreds of billions of dollars In new data centers. With the expansion of infrastructure, Demand for autonomous security systems It rises in parallel. The global security robotics market is expected to reach $19.18 billion in 2026, and double to $45.31 billion by 2033, representing an average annual growth rate of 13.1 percent.
The rationale is clear and straightforward. Securing modern infrastructure by relying on human personnel alone is increasingly inefficient: the volume of repetitive monitoring tasks is too high, and the demand for constant supervision is too great. On the other hand, robots can handle patrols and industrial inspections, including in environments that are difficult or dangerous for humans.
Meanwhile, the deployment of autonomous systems in public spaces has been increasing over the past few years. In October 2025, Dubai Police introduced an autonomous patrol robot In the Global Village entertainment complex. The robot moves autonomously through crowds, captures 360-degree video and sends it to the control center.
In January a Robot to direct traffic It first appeared at a busy intersection in Wuhu, China. In the UK, Nottinghamshire Police is Automated dog test In armed sieges and hostage-taking scenarios, the robot enters first to assess conditions while all decisions regarding force remain with the officers.
Across these different countries and systems, one principle remains constant: robots watch, people decide.
Why is the border located here?
At first glance, the idea that robots watch while humans act seems to contradict the general narrative put forward by many AI companies, which have long proposed a path toward fully autonomous systems. But in practice, current technology does not support this transition without causing unacceptable risks.
The linguistic models that underpin most modern intelligent systems do not have a firm understanding of the physical world. They do not understand reality in the human sense. they Works on probabilistic symbols and patterns Derived from large data sets. These systems clearly excel at tasks that exist entirely within structured code. But when we face unpredictable complexities in the real world, text is not enough. In these situations, models can “hallucinate,” confidently producing plausible but incorrect outputs. In the chatbot interface, this can be controlled as long as the output is reviewed. In infrastructure management or public safety, the consequences are much more serious.
Take, for example, a police robot that patrols the streets at three in the morning and encounters situations that cannot be predicted in advance based on training data. Someone lying on the sidewalk – is this an assault victim, an unconscious person, or a drunk person? On the surface, the visual signal may be identical, but the response required is completely different. Or take someone aggressively trying to open a car door – is he stealing the car or just trying to get into his own car?
Misinterpretation in these contexts can escalate into conflict, civil rights violations, or reputational crises for cities and operators.
The threshold of a self-driving car is an analogy
A useful analogy for what we are already striving to achieve is self-driving cars. It’s not enough for industry leaders like Waymo to show the public that their systems are, on average, no worse than human drivers. Organizers required demonstrable statistical superiority: Fewer accidents and fewer accidents At equal distances.
This threshold remains a matter of debate, but the principle is clear: the greater the potential harm caused by an error, the higher the level of proven safety. This is especially true for a robot, which may one day be granted the right to use force. If armed police robotic systems are to be introduced, they will have to demonstrate not only reliability comparable to that of a human officer, but also multifaceted superiority across all key metrics in real-world, not laboratory, conditions.
At the moment, we are still far from that threshold. Modern robotic police officers are intentionally unarmed and serve primarily as a substitute for patrol cars.
Responsible independence is the modern norm
The private sector has already learned the hard way that overestimating AI’s decision-making ability comes at a high cost. In 2023, Swedish fintech company Klarna – Laying off about 700 employees After introducing an AI-powered chatbot to handle their work, they were quietly rehired two years later.
Across all industries, companies that performed well in controlled demonstrations did so You encounter dirty dataNon-standard requests and hidden operational costs. Many of them continue to lose an estimated 40 percent of expected productivity due to manual correction of AI-generated errors.
As long as AI lacks a firm model of reality, and as long as hallucinations remain a regular feature rather than an exception, crucial decisions must remain under human control. Robots guarding data centres, autonomous patrols in Dubai, and mechanical police dogs should not give the impression that we are moving towards a world in which machines make decisions instead of humans. It points to something more real: the functional division of labor between humans and machines.
The authority to make final judgments must rest with responsible individuals. Only under this condition can progress remain stable and sustainable.
