Facing up to police facial recognition
Privacy advocates are rightfully raising questions about a police test of facial recognition on the streets of London this week. Not least because the tech isn’t particularly effective in spotting criminals while simultaneously having the potential to invade the public’s right to privacy as they go about their day.
The force has employed the tech – which scans people’s faces against a list of individuals of interest to the police – on numerous other occasions, including at the Notting Hill Carnival and a shopping centre in Stratford, east London.
This time, the tech will be attached to vans and run for eight hours on each day as the Met said it needed to put the tech through its paces for a longer time period.
However, use in a live environment is controversial because of poor success rates and a lack of governance and regulation.
In May, a Freedom of Information request from Big Brother Watch showed
the Met’s facial recog had a 98 per cent false positive rate.
The group has now said that a subsequent request
found that 100 per cent of the so-called matches since May have been incorrect.
I’m not aware of any evidence that other facial recognition systems deployed by law enforcement around the world (and apparently Taylor Swift
) are any better. China trumpets the success of its system, but maybe that’s just a placebo effect – tell people it works and they’ll be too scared to break the law.
Honestly, for all our objections, it’s very likely facial recognition will be just another part of law enforcement in a decade or so. Just as we quickly forgot the security cameras that follow us around city centres, there will quickly be enough support to keep facial recognition in use if it catches a widely hated criminal or three.
Whether data protection law can keep up is another question. But think about it – if Google and Facebook know everywhere you’ve been thanks to their apps, law enforcement squads are probably going to expect to have access to similar data as standard soon enough.