Who Controls the Robot? Human Control, AI Autonomy and the Question of Safety

Robots and AI systems are increasingly used in manufacturing, healthcare, transport and agriculture, raising a central question: who is responsible when a robot causes harm? This conceptual article examines the relationship between human control, AI autonomy and robotic safety. It distinguishes remotely controlled, autonomous and semi-autonomous robots and compares the risks of each. Human-controlled robots may be misused or operated incorrectly, autonomous systems may fail through technical faults or limits in their training, and semi-autonomous systems can blur responsibility between operators, developers, manufacturers and organisations. The 2018 crash of an automated test vehicle in Tempe, Arizona, as documented by the US National Transportation Safety Board, illustrates how responsibility can be shared between a human, a machine and an organisation. The article also sets out both sides of the debate on long-term AI risk and the term "superintelligence". It argues that safety depends not on choosing between human control and machine autonomy, but on matching autonomy to the consequences of failure, meaningful human oversight, rigorous testing, clear accountability, safety standards and education.