The police chose to trust the algorithm rather than their own eyes.
In the UK, another controversy over facial recognition technology has escalated into legal action. The case stems from the case of 26-year-old software engineer Elvi Chowdhury, who was arrested by police on suspicion of burglary in a city he had never visited.
In January, police officers arrived at the Choudhury family's home in Southampton, handcuffed him, and took him to the station. He was held in custody for nearly ten hours and released only around 2:00 a.m. The arrest was prompted by the results of an automated facial recognition system used by Thames Valley Police. The algorithm matched his photograph with CCTV footage from a £3,000 theft in Milton Keynes, approximately 100 miles from his home.
According to Chowdhury, the man in the video was noticeably younger, beardless, and with different facial features. He maintains that the difference in appearance is obvious, but it did not prevent the arrest. Thames Valley Police stated that the decision was made after a visual assessment by an officer, and the algorithmic match was considered only as a guide. The agency acknowledged that the error could have been due to bias in the technology, but emphasized that the arrest was not unlawful and was not a result of racial profiling.
British law enforcement agencies use an algorithm from the German company Cognitec, purchased by the Home Office. The system conducts approximately 25,000 searches monthly against a database of 19 million police photographs. According to a study commissioned by the same ministry, under certain settings, the false match rate for black people is 5.5 percent, for Asian people 4 percent, and for white people 0.04 percent. Police and Crime Commissioners have previously called these rates alarming and pointed to built-in bias. For more information on how these biometric systems work , see this article on SecurityLab.
Since December, Thames Valley Police has also been deploying real-time facial recognition technology in Oxford, Slough, Reading, Wycombe, and Milton Keynes. Since then, approximately 100,000 people have been covered by CCTV , resulting in six arrests.
Choudhury's photo ended up in the database due to a previous incident in 2021, when he was mistakenly detained after being assaulted while studying in Portsmouth. The case was closed without charges. Now, he fears the new photo increases the risk of repeated mistakes. He has filed a compensation claim against Thames Valley Police and Hampshire Police, the officers responsible for the arrest. Among the consequences, he cites family stress, attention from neighbors, and the inability to go to work the following day. Storing images in law enforcement databases directly affects personal data protection —an increasingly sensitive issue in light of new legislative requirements.
In December 2024, Biometrics and CCTV Commissioner William Webster expressed concern that police continued to retain images of people who had never been charged. Last month, South Wales Police paid compensation to a man who was also wrongly detained after being triggered by facial recognition.
The Home Office said guidance and training to reduce errors were being reviewed, and a new national face-matching system was being developed and would use an improved algorithm that had been independently tested.
In the UK, another controversy over facial recognition technology has escalated into legal action. The case stems from the case of 26-year-old software engineer Elvi Chowdhury, who was arrested by police on suspicion of burglary in a city he had never visited.
In January, police officers arrived at the Choudhury family's home in Southampton, handcuffed him, and took him to the station. He was held in custody for nearly ten hours and released only around 2:00 a.m. The arrest was prompted by the results of an automated facial recognition system used by Thames Valley Police. The algorithm matched his photograph with CCTV footage from a £3,000 theft in Milton Keynes, approximately 100 miles from his home.
According to Chowdhury, the man in the video was noticeably younger, beardless, and with different facial features. He maintains that the difference in appearance is obvious, but it did not prevent the arrest. Thames Valley Police stated that the decision was made after a visual assessment by an officer, and the algorithmic match was considered only as a guide. The agency acknowledged that the error could have been due to bias in the technology, but emphasized that the arrest was not unlawful and was not a result of racial profiling.
British law enforcement agencies use an algorithm from the German company Cognitec, purchased by the Home Office. The system conducts approximately 25,000 searches monthly against a database of 19 million police photographs. According to a study commissioned by the same ministry, under certain settings, the false match rate for black people is 5.5 percent, for Asian people 4 percent, and for white people 0.04 percent. Police and Crime Commissioners have previously called these rates alarming and pointed to built-in bias. For more information on how these biometric systems work , see this article on SecurityLab.
Since December, Thames Valley Police has also been deploying real-time facial recognition technology in Oxford, Slough, Reading, Wycombe, and Milton Keynes. Since then, approximately 100,000 people have been covered by CCTV , resulting in six arrests.
Choudhury's photo ended up in the database due to a previous incident in 2021, when he was mistakenly detained after being assaulted while studying in Portsmouth. The case was closed without charges. Now, he fears the new photo increases the risk of repeated mistakes. He has filed a compensation claim against Thames Valley Police and Hampshire Police, the officers responsible for the arrest. Among the consequences, he cites family stress, attention from neighbors, and the inability to go to work the following day. Storing images in law enforcement databases directly affects personal data protection —an increasingly sensitive issue in light of new legislative requirements.
In December 2024, Biometrics and CCTV Commissioner William Webster expressed concern that police continued to retain images of people who had never been charged. Last month, South Wales Police paid compensation to a man who was also wrongly detained after being triggered by facial recognition.
The Home Office said guidance and training to reduce errors were being reviewed, and a new national face-matching system was being developed and would use an improved algorithm that had been independently tested.
