They assessed emotions, looked for intruders and made sure the floors weren’t wet
Newly discovered documents show that over the past couple of years, Amazon's cameras and AI software have scanned the faces of thousands of rail passengers in the UK.
The surveillance system was tested at eight railway stations across the country, including London Euston and Waterloo terminals, and Manchester Piccadilly station. The trial was supervised by Network Rail, which manages the UK's rail infrastructure, with the aim of reducing crime by quickly alerting staff to safety incidents.
Network Rail itself did not respond to queries regarding the results of the testing and the current status of the system. But it said it takes the security of the rail network very seriously and uses a range of advanced technologies to protect passengers, colleagues and the rail infrastructure from crime and other threats.
Surveillance systems have been trained to automatically detect people stepping onto the tracks, potential criminals, and even antisocial behavior that includes running, shouting, skateboarding or smoking. Separate sensor tests were conducted on slippery floors, overflowing trash cans and clogged drains that could cause a mess.
Documents show that AI can use images of people to perform statistical analyzes of the age range and demographic characteristics of men and women. He was also able to analyze emotions such as happiness, sadness and anger by scanning facial expressions. It is unclear how widely emotion detection analysis has been used, but documents sometimes say that this use case should be “considered with greater caution,” and station reports say that “it is not possible to verify its accuracy.” .
Network Rail documents also claim that cameras installed at Reading station allowed police to speed up investigations into bicycle thefts because they were able to accurately identify the bikes in the footage.