Today's piece in the Guardian from David Davis MP, quoted below, is only the latest in an increasing tide of opinion pieces decrying the use of facial recognition technology, and (almost invariably) calling for an outright ban. Most of the writers seem to consider that, uniquely among the technologies currently being developed, facial recognition is so inherently malign that it simply cannot be permitted under any circumstances.
Such a position stems not only from a misunderstanding of the technology, but also from a misunderstanding of the applicable law.
While the reality is of course a good deal more sophisticated than this, automated facial recognition is at heart a camera and a database. No one sensibly suggests that police forces or private businesses shouldn't be able to ensure security through monitoring CCTV cameras. Nor would anyone credibly seek to argue that wrong-doers detected via those CCTV cameras should not be apprehended.
The concern is, or seems to be, that an automated system will make more mistakes than human operators. Now that may be true, at the moment, with the technology in its infancy. That is precisely why trials need to take place so that improvements to accuracy can be made. But in every field in which has been properly and rigorously trained it has started out being less capable than humans and ended up surpassing them. There is no reason to think that this would not be the case here.
So perhaps the concern is that the system could become too effective, that individuals, whether they had ever committed a crime or not, could be effortlessly tracked in every moment of their daily activities. That is a more reasonable concern - the prospect of this technology in the hands of a police state would be a real cause for alarm. But, crucially, it is a concern which takes no account of the laws already in force which prevent such a use of the technology, at least in this country and in the wider European Union.
This is a point which criticisms of the technology, and of the decision in the Bridges v South Wales Police case, completely misunderstand. Far from failing to tackle the conflict between this technology and civil liberties, the Bridges decision was a classic example of a judgment which carefully weighed the checks and balances inherent in human rights and data protection law and found that this particular trial, with all of the safeguards put in place around it, got the balance right. Fundamental in that decision were three elements, which will be important for any future use of the technology:
1. Publicity - the trial was conducted with associated advertising and notices which gave information both about the fact that the facial recognition technology was in operation, but also why it was being used.
2. Anonymity - the vast majority of images captured by the system were never accessed by a human. They were stored only for as long as was needed for the purposes of the technology, which meant that most of them were only retained for moments before the AI determined that they were not on the watchlist and deleted. Even those who were identified as a match were kept securely and for no longer than necessary.
3. Human supervision - this was said by the Court to be a critical factor. The technology did no more than highlight potential matches, which the human operators then reviewed and decided for themselves whether to act on.
Each of these factors were, moreover, elements that had been incorporated by design into the way in which the system had been developed and deployed - safeguards to meet the potential impact of the technology as assessed in the mandatory Data Protection Impact Assessment which must accompany any use of facial recognition technology.
None of this is to say that there is not a risk of the technology being mis-used, still less that there is not work to be done on inherently biased or discriminatory algorithms. But let us at least recognise that the position is not as bleak as it is currently being painted. Those crying out for safeguards around this new technology should not be so hasty to dismiss the decision of the High Court in Bridges. It is evidence that those safeguards are present, and that they are working.
Sadly, the high court in Wales did not grasp the conflict with civil liberties, recently ruling that a facial recognition trial by South Wales police was legal. The court made constant comparisons between facial recognition and the police’s use of fingerprints and DNA. But they are entirely different. Officers take fingerprints or DNA samples when they have a reasonable suspicion of a crime. They do not put a DNA or fingerprint scanner on the roadside and collect our data en masse.