The London Underground recently ran a year-long trial using AI surveillance technology. This trial saw the movements, behaviours and body language of thousands of passengers across the London transport system being monitored to identify potential threats and safety issues. The AI was trained through live CCTV feeds to pinpoint aggressive behaviours, instances of fare evasion, potential accidents, and the presence of weapons.
Transport for London carried out this test from October 2022 to September 2023, implementing 11 different algorithms at the Willesden Green Tube station. Although the local police force has used AI surveillance previously for public events, this marks the first time Transport for London has integrated AI with real-time video analytics to create alerts for immediate staff action. The trial resulted in around 44,000 alerts, nearly half of which were sent directly to staff for intervention.
Despite these numbers, the technology has faced significant criticism, particularly from human rights organizations. They consider it to be invasive and prone to errors and discrimination, especially in light of multiple ill-performing global AI surveillance projects, some of which have falsely associated crimes with individuals having darker skin tones.
The alerts generated during the London Underground trial were diverse: 26,000 alerts pertained to fare evasion, 59 to wheelchair users in a station lacking proper access, 2,200 for individuals crossing platform safety lines, 39 for people leaning perilously over platform edges and 2,000 for those sitting on benches for extended periods. Warning were also issued regarding aggressive behaviour, though this more nebulous category was difficult for the AI to reliably identify due to the lack of sufficient training data.
The overall aim of the AI system was to augment safety and security throughout the Underground. It was programmed to detect a range of scenarios, from illegal access and mobility assistance needs to identifying potential platform hazards. Officers even joined the tests, brandishing weapons within the CCTV field of view during off hours to enhance the AI’s responsive capabilities. However, there were still inaccuracies in the AI’s functionalities, such as erroneously detecting children as fare evaders.
Transport for London maintains that the ultimate goal is to create a safer and more efficient tube system for both the public and staff.
