Introduction
Rapid advances in technology are disrupting the balance of power between governors and the governed. Automated facial recognition technology (AFR) is not the only biometric surveillance technology that is in ascendency and posing challenges for law, but it is perhaps the one that has most captured the public imagination. AFR is an algorithmic technology. AFR algorithms are developed to locate a face within an image, measure the geometric features of the face (distance between eyes, width of mouth etc), and then ‘match’ the face to a previously stored image of the individual (usually stored on a watchlist or database), based on the strength of the correlation between geometric features.Footnote 1 It has numerous applications, such as automating border identity verification checks,Footnote 2 pupil registration in schools,Footnote 3 or enabling more effective photo sharing on social networking sites.
AFR is being utilised by law enforcement agencies across the world to fulfil various functions. In some jurisdictions with autocratic political regimes, AFR has been used for overt repression and persecution. One egregious example is the Communist Party of China, which has invested heavily in developing the infrastructure to pervasively monitor and control citizens. In 2015, Party officials called for an acceleration of public security video monitoring systems in order to achieve ‘systematic and dense coverage’ of all public areas.Footnote 4 In particular, the Party has targeted its Uyghur Muslim population in the Xinjiang region, installing AFR cameras, supplied by global surveillance manufacturer Hikvision, at the entrances to 967 mosques.Footnote 5 Most security checkpoints stationed along Xinjiang's major roads now employ facial recognition cameras of varying sophistication.Footnote 6 The technology is facilitating the Communist Party of China's authoritarian control of the region, where internment in ‘re-education camps’ and other human rights abuses of the Uyghur population have been well documented.Footnote 7
Chinese AFR surveillance infrastructure is also being purchased and utilised by Governments in developing nations, such as Uganda and Zimbabwe.Footnote 8 Other autocratic regimes are looking to AFR to strengthen controls on their citizens’ movements. In January 2020, Amnesty International reported that Russian authorities plan to operationalise a large-scale facial recognition system covering the whole Moscow Metro transportation network.Footnote 9 Whilst the use of AFR in these jurisdictions is not constrained by the checks and balances of jurisdictions with stronger legal human rights protections, the fast-paced expansion of experimental AFR in jurisdictions with broad commitments to liberal democracy, without prior parliamentary approval, raises legitimate questions concerning the effectiveness of the operation of any such checks and balances. Particularly, as to whether these checks and balances are robust enough to harness law enforcement to operate within the constraints of human rights law before experimental forms of surveillance technology are deployed.
In the US,Footnote 10 Australia,Footnote 11 and several EU Member States,Footnote 12 AFR has been used by police forces to monitor public spaces, and to retrospectively identify suspects by running CCTV images collected from a crime scene against police databases. Furthermore, the Covid-19 pandemic has prompted the use of facial recognition software to ascertain if someone is wearing a mask, the amendment of algorithms to be able to identify someone who is wearing a mask, and also AFR's combination with thermal imaging to determine if someone has a temperature, which is a common symptom of Covid-19. Such uses are evident in public spaces, workplaces and vulnerable/hotspot locations like aged-care facilities.Footnote 13
In England and Wales, police forces have used the technology for retrospective identification,Footnote 14 and to identify suspects in their custody.Footnote 15 AFR has also been used by police at a number of public gatherings to identify ‘persons of interest’ in real time. Owing to concerns over its impact on human rights, the police use of AFR has sparked public protestsFootnote 16 and counter-surveillance reactions, including the use of face coverings,Footnote 17 and the tactical use of lasers to obstruct AFR apparatus.Footnote 18 Civil liberties group, Liberty, has campaigned to ban the police use of AFR altogether, particularly in public spaces.Footnote 19 So far, these calls have been resisted by English lawmakers but, as we shall discuss, the regulation of police AFR has been, and continues to be, an evolutionary process. In September 2019, the High Court of Justice in R (Bridges) v The Chief Constable of South Wales, held that the law permits police the discretion to use AFR for various operational functions.Footnote 20 The Divisional Court found in the respondent's favour despite ‘the lack of a clear legislative framework for the technology’.Footnote 21 Indeed, the Protection of Freedoms Act 2012 provides a legal framework for two types of biometrics, DNA and fingerprints, but does not apply to other biometrics such as facial images, gait, or voice. In August 2020, the Court of Appeal allowed the claimant's appeal on the grounds that South Wales Police's (SWP) use of AFR was unlawful as it was not ‘in accordance with law’ for the purposes of Article 8(2) of the European Convention on Human Rights (ECHR), and the SWP had failed to carry out a proper data protection impact assessment (DPIA). SWP also failed to comply with the public sector equality duty. This was reportedly the first successful legal challenge to AFR technology use in the world.
This paper subjects the decisions in Bridges and, more broadly, the Government's position on the legality of police AFR use, to critical scrutiny. It will argue that the legal basis for police AFR surveillance is inadequate, and that the legal framework regulating the police use of this technology is too imprecise to protect citizens from abuse or arbitrariness. Drawing on the experience of other common law jurisdictions, and relevant socio-legal scholarship, the paper suggests that this framework and its interpretation in domestic courts transforms human rights protection into a tick-box exercise, foreclosing important political debate on the normative consequences of police AFR surveillance. English and Welsh police forces have been afforded too much discretion to ‘widen the net’ of biometric surveillance, without explicit democratic approval.
1. AFR in England and Wales
Since at least 2014, several police forces in England and Wales have trialled the use of AFR in several contexts. AFR is used in three broad ways:
(1) identity verification: a suspect is arrested but refuses to provide their name to police. Here, police could take a static ‘probe image’ of the individual's face. AFR software could then be used to verify the individual's identity by comparing the probe image against a database of images that the police control, or to which the police have access;
(2) retrospective or speculative identification: CCTV footage shows a suspected burglar leaving a property. A still of the suspect's face is used as a probe image and compared with a database of custody images (commonly known as ‘mugshots’). The AFR software generates a shortlist list of possible matches, and police arrest a suspect based on their place of residence being close to the crime scene and the strength of the AFR ‘match’;
(3) live AFR: a live deployment of AFR may be used to identify ‘persons of interest’ to the authorities as they traverse a vicinity of public space. Live AFR typically involves the deployment of surveillance cameras to capture digital images of members of the public, which are then compared with digital images of persons on a pre-assembled ‘watchlist’ of images the police have compiled for the purpose of the deployment.
The decision to use or not use AFR in a particular context is value-laden. Tensions between competing values and aims arise at different points in the development, deployment and maintenance of AFR, as with any other algorithmic system.Footnote 22 For example, the use of AFR for identity verification will be more accurate and safe than for a live AFR system, as it will involve the comparison of one clear and static image with another, also likely of high quality (a one-to-one comparison). Live AFR is likely to involve a ‘one to many’ comparison, that is, a target image with a database of subjects, or ‘many to many’ search, dataset to dataset. The live system may also bring into play human rights considerations that do not apply when utilising the first type (eg it may have a disruptive ‘chilling’ effect on public assemblies if used to monitor crowds at a protest or sporting eventFootnote 23). Likewise, speculative identification through AFR may raise issues pertaining to criminal procedure and fair trial rights that do not arise with other types of AFR, when these are used in a non-investigative capacity.Footnote 24
In England and Wales, several police forces have experimented with the use of AFR in numerous contexts. The most prominent trials have been conducted by Leicestershire Police, SWP, and the Metropolitan Police Service (MPS).Footnote 25 MPS have since gone beyond trialling the technology, adding AFR to their operational public surveillance arsenal.Footnote 26
Domestic forces have also collaborated with private organisations to engage in AFR surveillance activities. In August 2019, it was reported that several commercial landowners were using AFR surveillance on their publicly accessible land.Footnote 27 In September of the same year, MPS acknowledged that they had supplied images of individuals for one commercial landowner to use when deploying AFR at its site in the Kings Cross area of London.Footnote 28 In January 2020, the New York Times reported that a small AFR start-up company, Clearview AI, had sold its AFR tool to over 600 law enforcement agencies around the world, and a number of police forces based in the UK had accessed the Clearview AI database. The tool, which ‘scrapes’ and compares photos without consent from online platforms, is described as follows:
You take a picture of a person, upload it and get to see public photos of that person, along with links to where those photos appeared. The system – whose backbone is a database of more than three billion images that Clearview claims to have scraped from Facebook, YouTube, Venmo and millions of other websites – goes far beyond anything ever constructed by the United States government or Silicon Valley giants.Footnote 29
Many of Clearview AI's clients, including several UK and Australian police forces, had not previously disclosed – or indeed had denied – their use of the app to the public, and had no internal guidance or policies regulating the circumstances in which they might use the app.Footnote 30 The Information Commissioner's Office and its Australian counterpart have opened a joint investigation into Clearview AI's personal information handling practices.Footnote 31
The rise of AFR has not been frictionless. Civil liberties organisations, such as Liberty, various academics (including the authors of this paper) as well as the Home Office Biometrics and Forensics Ethics Group, have raised concerns that the police use of AFR lacks a firm legal basis, and poses an acute threat to various human rights, including the rights to privacy, free association, and freedom from discrimination.Footnote 32 In July 2019, the House of Commons Science and Technology Select Committee called on the Government to issue a moratorium on live AFR ‘until a legislative framework has been introduced and guidance on trial protocols, and an oversight and evaluation system, has been established’.Footnote 33 Their concerns were, it seems, well founded. In the same month, an independent report of the MPS's trial of live AFR identified several issues arising from the governance of the trials, pertaining to the very low number of arrests made in comparison to numbers scanned by the system; its operational use by the Met; and the legal basis relied on for the trials.Footnote 34 Perhaps owing to concern that this independent evaluation did not produce the desired results, the MPS produced their own evaluation report, which took a more positive view of the trials and concluded that AFR will ‘stop dangerous people and make London safer’.Footnote 35 The document made no reference to the independent evaluation study, and did little to address its main findings.
The Information Commissioner, Biometrics Commissioner and the Surveillance Camera Commissioner have all expressed concerns about the rapid rise of AFR in a policing context, too.Footnote 36 In February 2020, almost six months after the Divisional Court judgment in Bridges had been handed down, Lord Clement Jones sponsored a Private Members Bill, which would prohibit the use of AFR technology in public places and to provide for a review of its use.Footnote 37 At the time of writing, no date has been announced for the Bill's second reading.
Despite this widespread criticism and political resistance, the rise of AFR in policing continued apace. SWP continued to trial AFR for both live surveillance and identity verification. MPS have used live AFR at numerous shopping centres and public places in 2020, and several other forces are reported to have used AFR surveillance, without publicising their trials.Footnote 38 Parliament has not introduced any specific laws relating to AFR. The police have maintained that the legal basis regulating its proper operational limits lay in the DPA 2018, the Surveillance Camera Code of Practice, and relevant common law and human rights principles. As indicated above, these arguments were put to the test in the Bridges case, when a Divisional Court considered an application for judicial review of the legality of SWP's use of AFR.Footnote 39
2. R (Bridges) v Chief Constable of South Wales Police
SWP is the national lead on AFR, having received a £2.6 million government grant to test the technology.Footnote 40 A Cardiff resident, Mr Bridges (described as a civil liberties campaigner) had challenged the legality of SWP's general use and two particular deployments of AFR on the grounds that this was contrary to the Human Rights Act 1998, data protection legislation, and that the decision to implement it had not been taken in accordance with the Equality Act 2010. The Divisional Court rejected this application.
It is worth highlighting some of the dimensions of the SWP initiative, in order to illuminate the reasons behind the Divisional Court's refusal of judicial review, and the later overturning of this decision on appeal. In April 2017, SWP began a trial of automatic AFR with subsequent national rollout in mind. The trial comprised two pilots, one of which was known as AFR Locate and the other known as AFR Identify. The judicial review proceedings concerned AFR Locate; a form of live AFR. This involves the processing of digital images of members of the public taken from live CCTV feeds, and the comparison of these images with biometric information of individuals on a watchlist compiled specifically for the purpose of the deployment of AFR. SWP took steps to inform members of the public that AFR Locate was being used at a particular event or area by posting on social media and putting signs up in the vicinity of the trial.
In terms of human rights, the Divisional Court concluded that while the use of AFR Locate engaged the ECHR Article 8 (privacy) rights of the members of the public whose images were taken and processed, there was no violation of Article 8 as SWP's use of AFR was ‘in accordance with the law’ and was ‘necessary in a democratic society’, and in pursuit of the legitimate aim of preventing or detecting crime, for the purposes of Article 8(2). On the legality point, AFR use was deemed to be within the police's common law powers so that there is currently no need to legislate to permit its use and it was not ultra vires, at least as practised in the SWP trials. Moreover, those actions were subject to adequate legal controls, contained in data protection legislation, statutory codes of practice, and SWP's policies. The pilots were legally justified; AFR Locate was deployed only for a limited time, and for specific and limited purposes. Furthermore, unless someone's image matched that on the watchlist, all data were deleted immediately after having been processed. The CCTV feed is retained for 31 days in accordance with the standard CCTV retention period, and data associated with a match is retained within AFR Locate for up to 24 hours. In its necessity analysis, the Divisional Court found that the benefits of live AFR were potentially great, as serious offenders might be apprehended, and the impact on Mr Bridges was relatively minor as he was not stopped by the police, and so the use of AFR was proportionate under Article 8(2).
As for the data protection claims, the Court determined that the collection and processing by SWP of images of members of the public constituted collecting and processing of their personal data, notwithstanding that they might not be identifiable by name. Such processing of personal data was deemed to be lawful and to comply with the conditions in the DPA 2018. The Court was also satisfied that SWP had complied with the requirements of the public sector equality duty.
Mr Bridges sought leave to appeal the Divisional Court's decision on the following five grounds:
(1) the Divisional Court erred in concluding that the interference with the appellant's Article 8 rights occasioned by SWP's use of AFR was ‘in accordance with the law’ for the purposes of Article 8(2) ECHR;
(2) the Court made an error of law in assessing whether SWP's use of AFR constituted a proportionate interference with the appellant's Article 8 rights;
(3) the Divisional Court erred in holding that SWP's Data Protection Impact Assessment complied with the DPA 2018, s 64;
(4) the Divisional Court erred in declining to reach a conclusion on whether SWP has in place an appropriate policy document within the meaning of the DPA 2018, s 42 (taken with s 35(5)), which is a prerequisite for lawful data processing in this context;
(5) the Divisional Court made an error of law in concluding that SWP had complied with the Public Sector Equality Duty in the Equality Act 2010, s 149, given that its approach to the equalities implications of AFR is ‘demonstrably flawed’ as it failed to recognise the risk of indirect discrimination.Footnote 41
On the first ground, the Court of Appeal ruled that the Divisional Court did err in its finding that the measures were ‘in accordance with the law’. The Court of Appeal did not revisit the issue of whether or not the use of AFR was ultra vires the SWP, holding that the police had long used overt surveillance techniques such as overt photography that were ‘undoubtedly’ in accordance with the law.Footnote 42 Instead, the Court engaged in a holistic analysis of whether the framework governing SWP's use of live AFR was reasonably accessible and predictable in its application,Footnote 43 and sufficiently prescribed to guard against ‘overbroad discretion resulting in arbitrary, and thus disproportionate, interference with Convention rights’.Footnote 44
The Court of Appeal rejected the contention that for the police to use AFR they needed some statutory authorisation, but accepted, applying a relativist approach, that more would be required by way of safeguards for AFR than for overt photography as the former was a novel technology that involved the automated processing of sensitive data.Footnote 45 However, unlike the Divisional Court, the Court of Appeal was not satisfied that SWP's use of live AFR was sufficiently regulated by the combination of the DPA 2018, the Surveillance Camera Code of Practice and SWP's local policies. In particular, the legal framework left too much discretion in the hands of individual officers to determine who was to be placed on the watchlist, and where AFR could be deployed.Footnote 46 Thus, the framework did not sufficiently set out the terms on which discretionary powers in these areas could be exercised and for that reason they did not have the necessary quality of law. The Court held that a police force's local policies could constitute relevant law in this context, provided they were published and would not leave too much discretion to individual officers. This finding is significant. In short, if SWP were to amend their own policies in such a way that the criteria for (i) who could be put on a watchlist, and (ii) where AFR could be used, were more narrowly circumscribed, this could be all that is needed to satisfy the ‘in accordance with the law’ limb of Article 8(2).
The appeal did not succeed on the second ground. Here, the Court held that SWP's use of AFR was a proportionate interference with Article 8 rights, and as such was ‘necessary’ and ‘in pursuit of a legitimate aim’ under Article 8(2). The appellant submitted that the Divisional Court fell into error as a matter of approach when addressing the question of proportionality, because it conducted a weighing exercise with one side being the actual and anticipated benefits of AFR Locate and the other side being the impact of AFR deployment on the appellant. The appellant argued that ‘as a matter of common sense’, account needs to be taken of the interference with the Article 8 rights not only of the particular appellant but all other members of the public in the vicinity of SWP's AFR deployments.Footnote 47 The Court of Appeal rejected this argument as the focus of the appeal was the impact upon the appellant alone, and, in any event, the impact on the Article 8 rights of other members of the public was as negligible as it was on the appellant. In the Court's words, ‘An impact that has very little weight cannot become weightier simply because other people were also affected. It is not a question of simple multiplication’.Footnote 48
On ground three, the Divisional Court erred in finding that SWP provided an adequate DPIA as required by DPA 2018, s 64. The ‘inevitable consequence’ of the DPIA being written on the basis that Article 8 is not infringed is that it failed to adequately assess the risks to the rights and freedoms of data subjects or include the measures envisaged to address the risks arising from those deficiencies as required by DPA 2018, s 64(3)(b) and (c). However, on the fourth ground the Court of Appeal held that it was ‘entirely appropriate’ for the Divisional Court not to reach a conclusion as to whether SWP had in place an ‘appropriate policy document’ within the meaning of DPA 2018, s 42, as the deployments at issue took place before the DPA 2018 was enacted.Footnote 49
Finally, the appeal succeeded on the fifth ground. The Court of Appeal held that the SWP never had due regard to the need to eliminate discrimination on the basis of sex and race. The Court of Appeal found that, whether or not the facial recognition software used by SWP is biased and creates a greater risk of false identifications among certain demographic groups, the police breached their positive duty to have due regard to the need to eliminate such discrimination; the ‘SWP have never sought to satisfy themselves, either directly or by way of independent verification, that the software program in this case does not have an unacceptable bias on grounds of race or sex’.Footnote 50
This issue of bias is contentious. It is clear that AFR varies in terms of accuracy and reliability in matching individuals to images, depending on gender, age, skin colour, etc, and such embedded biases may compound existing biases in policing. The National Institute of Standards and Technology (NIST), a subgroup of the US Federal Department of Commerce, has provided technical evaluation of over 100 commercially available facial recognition algorithms as part of its ‘Facial Recognition Vendor Tests’ (FRVT). They measure the accuracy of facial recognition software algorithms in ‘one-to-one’ (image verification) and ‘one-to-many’ (database search) contexts. The performance of face recognition systems can vary relative to the gender, ethnicity and age of the individuals targeted.Footnote 51 Part 3 of the National Institute of Standards and Technology's tests focused specifically on demographic effects on the performance of 189 commercially available facial recognition algorithms. It found that many of the algorithms varied in performance across different demographic groups, and that the part of the world in which the algorithm was developed could have a significant impact on its performance.Footnote 52 For example, algorithms developed in the US tend to have higher false positive rates for West and East African and East Asian people in one-to-one matching, whereas for a number of algorithms developed in China this effect is reversed, with low false positive rates on East Asian faces.Footnote 53 For ‘one-to-many’ matching, the test found that African American females were subject to high rates of false positives. This is significant because a false positive match on a ‘one-to-many’ search could put an individual at risk of being subject to scrutiny by authorities as a result of an incorrect match against a database. The Court of Appeal's finding on this fifth ground is significant and welcome; it will have implications stretching to the use by public authorities of algorithmic technologies generally. The Court of Appeal has made clear that public authorities have a positive duty to take measures, such as independent verification, to ensure that the technologies they use for processing sensitive personal data do not produce unacceptable demographic bias.
Bridges was the first ever legal challenge to the use of AFR. The Court of Appeal has underlined that the police do not enjoy boundless discretion to deploy biometric surveillance on the general public as they traverse publicly accessible space. In the aftermath of the Court of Appeal's decision in Bridges, the response of both MPS and SWP seems to suggest that judgment will not serve as the end of a legal saga on the limits of police AFR surveillance. SWP have indicated that they will not appeal the Court of Appeal's decision and in published comments following the judgment, stated:
There is nothing in the Court of Appeal judgment that fundamentally undermines the use of facial recognition to protect the public. This judgement will only strengthen the work which is already underway to ensure that the operational policies we have in place can withstand robust legal challenge and public scrutiny.Footnote 54
In MPS's own reaction, they gave clear indication that the judgment would not present significant obstacles to their own use of live AFR, emphasising that their own live AFR operations could be distinguished from SWP's trials; the main differences being that the MPS's use of this technology ‘… is intelligence-led, and focused on helping tackle serious crime, including serious violence, gun and knife crime, child sexual exploitation and helping protect the vulnerable’. As well as being more targeted in its use, MPS also claimed that their AFR operations are supported by their own bespoke guiding policy documents and ‘the latest accurate algorithm’.Footnote 55
The police commitment to use AFR has seemingly held firm. The police are considering how, and not whether, their AFR operations and policies can be brought into alignment with the requirements of law post-Bridges. This should not be too surprising, given that both SWP and the MPS have invested considerable resources in their AFR infrastructure to date.
Although the outcome of Bridges may appear at first glance to deal a heavy blow to the police's ambitions to continue to develop and deploy AFR, owing to the way the case was decided, the police may be able to make their AFR operations legally compliant through minor procedural amendments. For example, the Court of Appeal left open the possibility that an internal police policy document could be brought into accordance with the law for Article 8 purposes if it limited the discretion of individual officers as to who can go on a watchlist and where AFR can be used. SWP could clear this low hurdle by making tweaks to their own internal policies, despite the absence of a positive legal basis for the police use of AFR beyond general common law powers. The Court of Appeal also found that the use of AFR by SWP was proportionate in the face of reasoned doubts about its operational utility and lingering concerns about the human rights implications of scanning hundreds of thousands of people to yield comparatively few arrests.Footnote 56
In what follows, we assess the state of the regulation of police AFR surveillance under English law, following the Bridges judgment. Building on our prior work regarding the threats posed to human rights by AFR surveillance, and English law's response to the use by police of this technology,Footnote 57 and drawing on the experience of other jurisdictions and relevant policing jurisprudence and literature, we examine the ‘key moments’ in the Bridges judgments, where the legal conditions necessary for the continuing expansion of experimental police surveillance have been embedded in the English legal framework by the senior courts. Through this analysis, we suggest that this framework of broad, overarching safeguards still opens up vast operational discretion for the police to set the limits of ‘lawful’ use of an intrusive surveillance technology; the human rights and data protection framework functioning less as a barrier to government power, and more as a tool of post hoc legitimation, and an instrument of its expansion without prior democratic approval.Footnote 58 Thus, although the regulatory framework for police AFR surveillance is based on a discourse of respecting human rights, this discourse serves to provide cover for the expansion of pervasive surveillance in practice.Footnote 59
3. The legal basis for police AFR surveillance
Unlike in several continental European jurisdictions, there is no exhaustive code of criminal procedure providing explicit textual basis for the exercise of potentially coercive police powers.Footnote 60 Instead, the legal basis for the exercise of police power in England and Wales is multifaceted; it consists of statute, delegated legislation, and common law with each piece of this moveable mosaic shifting into its proper place depending on the power under exercise. For example, to compile a watchlist of images of offenders, the police will typically be able to rely on explicit statutory powers to collect and use of custody images, such as those contained in section 64A of the Police and Criminal Evidence Act 1984. Other statutes, such as the Data Protection Act 2018 and the Protection of Freedoms Act 2012 regulate aspects of AFR use by public and private authorities, providing for public notification requirements and retention limits, for example. However, under English law there is no explicit textual provision for the police use of AFR. Without explicit authorisation for the use of AFR from the legislature in the form of an authorising statutory provision, the police must find some other explicit or implicit power or liberty supporting their use of this technology.
For many of their day-to-day activities, the constable can rely on the same residual liberties that all citizens enjoy to do anything that is not expressly forbidden by law.Footnote 61 This residual liberty has its roots in the historical evolution of the police constable as a ‘citizen in uniform’ patrolling the streets for the benefit, and with the consent, of his fellow citizens.Footnote 62 In Malone v Metropolitan Police Commissioner, this residual liberty was held to extend very far indeed, covering the activities of the Post Office when they, acting on the request of the Metropolitan Police, tapped the telephone line of the plaintiff. Despite the absence of any statutory basis for the tapping of the telephone line by the authorities at the time, Sir Robert Megarry VC held that the tapping was not unlawful, as there was no positive legal right to immunity from this activity, observing:
England, it may be said, is not a country where everything is forbidden except what is expressly permitted: it is a country where everything is permitted except what is expressly forbidden.Footnote 63
The ‘citizen in uniform’ conception of the constable and the extent of the residual liberty he or she enjoys are both contestable,Footnote 64 but it is clear that this residual liberty no longer extends to cover intrusive and covert surveillance activities such as wiretaps and other communication interceptions. In Malone v United Kingdom, the European Court of Human Rights (ECtHR) held that the UK Government's reliance on this residual liberty did not satisfy the ‘in accordance with the law’ limb of Article 8(2).Footnote 65 By the time the judgment was delivered, the Government had already placed the regulation of public authority wiretapping on a statutory footing in the Interception of Communications Act 1985.
The idea that the police could depend on this residual liberty to engage in overt surveillance, such as recording and monitoring individuals in public spaces, has persisted. In Murray v United Kingdom, the first applicant was detained by the British Army in Belfast during the Northern Ireland conflict and taken to a screening centre. She argued, inter alia, that her Article 8 rights had been violated as she was photographed by the British Army, without her knowledge and consent, and the photographs were kept on record along with personal details about her, her family and her home. The domestic courts had dismissed her claim that the taking and retention of her photograph in these circumstances was in any way actionable, with the Court of Appeal in Northern Ireland finding:
The act of taking the photograph involved nothing in the nature of a physical assault. Whether such an act would constitute an invasion of privacy so as to be actionable in the United States is irrelevant, because the [first applicant] can only recover damages if it amounts to a tort falling within one of the recognised branches of the law on the topic. According to the common law there is no remedy if someone takes a photograph of another against his will.Footnote 66
Thus, the actions of the British Army were said to be lawful because they were not legally forbidden. The ECtHR rejected the applicant's contention that the measures lacked a legal basis such as is required for the measures to be ‘in accordance with the law’ under Article 8(2). The Strasbourg Court ruled that the common law provided sufficient domestic legal basis.Footnote 67
Then in R (Wood) v Commissioner of Police for the Metropolis, the Court of Appeal considered the legal basis for overt police photography surveillance.Footnote 68 The claimant – a political campaigner – was photographed by the police at a protest outside the Annual General Meeting of a company connected to the arms trade. A majority of the Court held that the police had violated the claimant's Article 8 rights because the collection and retention of the images was disproportionate in the circumstances. The Court did not reach a decisive view on whether the common law provided adequate legal basis for the activities of the police to be ‘in accordance with the law’. Although obiter, Laws LJ observed, ‘the requirement of legality is in my judgment satisfied by the general common law power referred to in Murray’.Footnote 69
This line of authority seems to indicate that the police do not require a positive legal power to engage in overt surveillance operations. However, in its more recent case law the ECtHR has been more exacting in its analysis of the quality of the legal basis underpinning surveillance measures, including the use of photography.Footnote 70 It is noteworthy in this regard that Murray v United Kingdom predates the enactment of the Human Rights Act 1998 which, under section 6, places a direct obligation on the police ensure that all of their activities are compatible with Convention rights. In any event, the police use of AFR, even in public spaces, can be distinguished from the police collection and retention of photographs of the applicants in Wood and Murray, because AFR involves biometric data processing. This point was recognised by both the Divisional Court and Court of Appeal in Bridges.Footnote 71 It is perhaps out of sensitivity to these developments in the legal and technological landscape that the police have tended in recent challenges to their use of overt surveillance to rely on positive powers to engage in various activities in the fulfilment of their basic common law duties, to justify the use of overt surveillance technologies.Footnote 72
In Bridges, the Divisional Court rejected the claimant's contention that the police could not rely on their common law powers to use AFR and, as such, the use of AFR was ultra vires the SWP. Haddon-Cave LJ and Swift J relied on the following passage from Rice v Connolly:
[I]t is part of the obligations and duties of a police constable to take all steps which appear to him necessary for keeping the peace, for preventing crime or for protecting property from criminal damage. There is no exhaustive definition of the powers and obligations of the police, but they are at least those, and they would further include the duty to detect crime and to bring an offender to justice.Footnote 73
Drawing on Wood and the Supreme Court judgment in R (Catt) v Association of Chief Police Officer, the Divisional Court held that this general power of the police covers the use, retention, and disclosure of imagery of individuals for any of the duties articulated in Rice.Footnote 74 In Catt, the applicant unsuccessfully argued that the retention by police of information about his attendance at several protests against the arms trade, which included written notes of his activities and a photograph, violated his Article 8 rights. Lord Sumption, in the majority, held that the power of the police to engage in surveillance of this kind lay not in the residual liberty of all citizens to do anything that is not forbidden, but rather in the general common law powers of the police officer to fulfil his basic duties: ‘At common law the police have the power to obtain and store information for policing purposes, ie broadly speaking for the maintenance of public order and the prevention and detection of crime’.Footnote 75
These powers, in Lord Sumption's view, did not extend to cover what he described as ‘intrusive’ methods of obtaining information, such as entry upon private property or acts of physical coercion, that are normally regulated by statute or other more narrowly prescribed sources. However, they were ‘amply sufficient’ to cover the obtaining and storage of photographs.Footnote 76 On this basis, the Divisional Court in Bridges classified live AFR as ‘non-intrusive’ and thus falling within the scope of the common law powers of the police. The Divisional Court ruled that the distinction turned on whether there was a physical intrusion with a person's rights vis-à-vis his or her home or interference with his or her bodily integrity.Footnote 77 The Court of Appeal upheld this aspect of the Divisional Court's ruling, citing Lord Sumption's observations in Catt with approval. The issue for the Court of Appeal was not that there was no adequate domestic legal basis providing for the use of AFR, but rather that this basis did not meet the quality of law requirements to be ‘in accordance with the law’.Footnote 78 It seems that only forms of ‘physical’ intrusion, such as DNA sampling and bugging private property, fall beyond the scope of the common law powers of the police, thus requiring a more narrowly prescribed legal basis.
This is a significant finding, as it permits the police to use new overt surveillance technologies like AFR operationally, without Parliament authorising this use. It puts the police collection and processing of biometric data by AFR in a separate category to other forms of biometric surveillance, such as arrestee DNA and fingerprint collection and comparison. These tend to require ‘physical intrusion’ and, as such, have a statutory legal basis under the Protection of Freedoms Act 2012. The effect of the Divisional Court's interpretation of these cases is that statutes are only relevant in so far as they place limitations on how AFR is used by the police. Thus, the police are free to trial and use these new technologies in the absence of the democratic mandate that legislation passed by Parliament provides. If this is the case, there is no need for police AFR use to be approved by our elected representatives, usually following robust debate on the implications of this use, and consideration of expert evidence scrutinised by Select Committees. The decision to use AFR is a matter for police to decide for themselves; their discretion on this matter is fettered only by the limits of their common law powers which, as the Court in Bridges acknowledged, are expressed in ‘very broad terms’.Footnote 79 It is noteworthy that in Catt v United Kingdom, the ECtHR expressed concern that the collection and retention by police of personal information taken from public space in that case ‘did not have a clearer and more coherent legal base’ before concluding that the retention violated Article 8 ECHR.Footnote 80
That said, it is difficult to fault the Divisional Court and Court of Appeal's interpretation of recent domestic authorities in Wood and Catt, which, as a matter of stare decisis, were binding on them in Bridges. These authorities do suggest that the general common law powers of the police set out in Rice extend to the collection, use, retention and dissemination of facial images. The problem is that, in interpreting the common law powers of the police so broadly, these authorities may have sent the law down a wrong path.
There is clear domestic authority for Lord Sumption's observation in Catt that physical acts which would otherwise constitute a technical assault or trespass fall far outside the scope of the common law powers of police.Footnote 81 However, it does not necessarily follow that the powers conferred upon police by their common law duties extend to cover all acts that are not ‘physically’ intrusive in the sense described in Bridges. For example, in Rice, the appellant successfully argued that the offence of obstruction of justice was not made out in circumstances where he merely refused to provide his name or other assistance to a police constable in the course of his investigation into a series of breaking offences. Lord Parker CJ held that police constables have a duty to take steps which appear necessary to prevent and detect crime. However, as Aston notes, this finding was ‘categorical’; the power of the constable did not extend to taking all steps he considered necessary for the prevention or detection of crime.Footnote 82 Indeed, Lord Parker CJ held that there are clear limits on the common law power (beyond those that interfere with the individual's physical home or bodily integrity), one being that citizens are not under a general legal duty to assist the police by providing them with information, and police cannot rely on their common law powers to demand such assistance. This was the unambiguous ratio of Rice.
We submit that other forms of non-physical coercion of citizens (such as verbally badgering or harassing an individual in the street) would also fall beyond the common law powers of the police, even if they involve no physical restriction of the individual's movement. None of the judgments of senior courts prior to Bridges drew the boundary of the common law powers so bluntly. Historically, domestic courts have tended to focus on whether the conduct of a police officer constitutes a significant interference with the individual's liberty or property generally, and not on the existence of physical contact. Thus, in the famous case of Collins v Wilcock, Goff LJ held
A police officer has no power to require a man to answer him, though he has the advantage of authority, enhanced as it is by the uniform which the state provides and requires him to wear, in seeking a response to his inquiry. What is not permitted, however, is the unlawful use of force or the unlawful threat, actual or implicit, to use force.Footnote 83
Here, the Court recognises that, as well as physical intrusions, excessive demands to answer questions, including the use of actual or implicit threats, fall outside of the common law powers. Indeed, some forms of physical touching will fall within the police common law powers, as in Donnelly v Jackman, where a police officer repeatedly tapped a man on the shoulder in order to get his attention for the purpose of making enquiries about an offence which the officer had cause to believe the defendant might have committed.Footnote 84 Even though this case involved making physical contact with the defendant, the Court focused on the degree of intrusion into the person's liberty and not on whether the type of intrusion involved trespass onto his physical realm. As Talbot J stated, ‘it is not every trivial interference with a citizen's liberty that amounts to a course of conduct sufficient to take the officer out of the course of his duties’.Footnote 85
The focus in Donnelly, on the significance of the intrusion with the person's liberty to use public space unmolested, seems better equipped for the information age than an arbitrary focus on whether the means used by police to obtain personal information occasioned physical contact. It is one thing to hold that the common law power to prevent crime and bring offenders to justice is good enough for the police to ask for identifying particulars of a person seen in the vicinity of reported criminality (even though the person may be under no legal obligation to comply with the request). It is quite another for this same power to support the use of myriad biometric and/or algorithmic technologies, which facilitate the use and collection of ever-more sensitive personal information by public authorities. AFR enmeshes physical and informational forms of surveillance by collecting information from the physical body of the person (albeit without occasioning physical contact) and breaking this down into an information structure, which can then be processed. The distinction for fleshing out the scope of the common law powers of the police, between physical and informational intrusions, does not seem to recognise that physical contact is a poor barometer for gauging the intrusiveness of a surveillance measure. The common law powers of the police enunciated in Rice have been extended too far.
(a) Proportionality, balancing and ‘tick-box’ human rights protection
One significant and unresolved issue with the English legal framework as it relates to AFR is that it leaves it largely to the police to decide whether the use of AFR (and thus further biometric technologies as they emerge) is ‘proportionate and necessary’, according to the precepts of the common law necessity test, Article 8(2), and the DPA and so on. The Appeal Court in Bridges considered the question of proportionality, and noted that, strictly speaking, it was unnecessary to do so.Footnote 86 Regrettably, this assessment was formalistic, not internally consistent, and unduly brief.
As indicated in the account of the case above, Mr Bridges submitted that the Divisional Court had erred when examining the ‘cost’ side of proportionality by taking into account the impact of AFR on him alone, and that ‘as a matter of common sense’ account should be taken of the interference with the Article 8 rights of all other members of the public present at the two deployments also.Footnote 87 This contention was rejected by the Appeal Court, which held that the challenge was to a ‘very specific deployment of AFR Locate on two particular occasions’. To come to this conclusion, it drew on the precise terms of the Statement of Facts and Grounds that this Appellant's Article 8 rights had been violated. It is regrettable that the ground on proportionality was not framed in more general terms. Indeed, both the Court of Appeal and the Divisional Court themselves framed the appeal as a review of the legality of SWP's ‘ongoing’ trial of AFR. Moreover, one can question how the proportionality of future use could be assessed adequately through reference to the impact it has on Mr Bridges solely. We suggest a wider lens should have been adopted in term of scrutinising proportionality, given the impact on other individuals in these deployments and beyond.
The Court of Appeal differentiated Bridges from challenges to a ‘general measure, for example a policy or even a piece of legislation’, such as in Tigere where the Supreme Court considered eligibility for student loans for English residents.Footnote 88 In such challenges to a ‘general measure’, it may be appropriate for the Court to assess the balance between the impact on every person who is affected by the measure and the interests of the community.Footnote 89 That was not the interpretation in Bridges. The Court could have considered this use of AFR a ‘general measure’, affecting as it did anyone in particular public spaces in Cardiff during certain times. That said, states are afforded a wider margin of appreciation when it comes to measures of general economic or social strategy,Footnote 90 so this would not affect the ultimate determination. Even without such an interpretation, the Court could have considered more fully the intrusion on every person subject to the deployment of the technology, as well as on the public en masse.
As well as limiting the breadth of the impact, the level of the perceived intrusion was also downplayed by the Court. The effect on members of the public who were in an analogous situation to Bridges was ‘as negligible as the impact on the Appellant's Article 8 rights’, and this impact of ‘very little weight cannot become weightier simply because other people were also affected’. The Court stated that this was not a question of simple multiplication, but rather an exercise of judgement.Footnote 91 There is little elaboration of what this judgement entailed. Moreover, in terms of consistency, it is hard to reconcile this statement about negligible impact with the holding of the Divisional Court that Article 8 was engaged by AFR technology, the comparison with fingerprints and DNA, and the description that AFR enables the extraction of unique information and identifiers of ‘an intrinsically private character’.Footnote 92 The public visibility of a person's face was not regarded as detracting from this, and Article 8 was engaged even though retention of the data was momentary. Furthermore, the Court of Appeal approved one of the grounds of appeal which stated that the DPIA had proceeded incorrectly on the basis that Article 8 was not engaged nor ‘more accurately was not infringed’.Footnote 93
It is difficult to see how Article 8 could be infringed just negligibly by AFR. That said, this is a common approach of domestic courts in surveillance cases, whereby the interference is characterised as minor or of negligible impact, before the benefits are trumpeted.Footnote 94 Courts seem to overlook the reality of overt and biometric surveillance as a preventive policing strategy geared towards the effective management of populations, and instead focus on individual (and physically intrusive) liberty restriction. This is a partial and questionable approach. This approach also seems out of step with ECtHR jurisprudence, which has consistently held that the recording and processing of an individual's photographic data is protected by Article 8. In Reklos v Greece, the ECtHR held
A person's image constitutes one of the chief attributes of his or her personality, as it reveals the person's unique characteristics and distinguishes the person from his or her peers. The right to the protection of one's image is thus one of the essential components of personal development and presupposes the right to control the use of that image.Footnote 95
The ECtHR has held that where a record of an individual's image is generated from CCTV cameras monitoring publicly accessible space, and retained by a public authority,Footnote 96 this can engage Article 8. AFR surveillance goes further than merely collecting and storing photographic images through CCTV surveillance as it involves the biometric processing of the facial geometry of those scanned. This enables public authorities to subject those who traverse public space to an automated identification process in real-time, without their knowledge or consent. Thus, AFR surveillance enables the police to go further in transgressing social norms governing the flow of information about individuals as they occupy public space than CCTV surveillance or other forms of overt photography.Footnote 97
The concept of proportionality is embedded in human rights jurisprudence and discourse, though it is not uncontroversial.Footnote 98 In this specific instance, the analysis in Bridges on proportionality overall was perfunctory, and did not really come close to grappling with the politically contentious issues at the heart of the debate on AFR, and its wider social effects, beyond the impact on the individual claimant. It might well be argued that a court is not best placed to engage in this sort of in-depth normative analysis, which is a fair remark. But, as Parliament has declined to regulate AFR in any meaningful way, and indeed is not required to do so, the decision of when and where AFR is proportionate is left largely in the hands of the police, who are institutionally ill-suited to striking a fair balance between crime control and human rights impacts. Although process-based judicial review aims to harness the legislature and police to broader human rights principles, this is inadequate because it entails a post hoc consideration where the role of the court is to engage in a limited review of AFR surveillance, grounded in broad common law compliance and a brief necessity analysis, and not to draft legislation on the appropriate limits of AFR.
Moreover, the full spectrum of harms cannot be captured through orthodox legal analytical lenses regarding human rights and data protection, which are focused on the impact on individual rights against state/data controller. The impact of the police use of AFR is greater than the impact it has on any one individual subject to AFR processing. It shifts the balance of power between state, who are using the products of private companies, and the public in favour of the former. In essence, law is not enough.
(b) Some comparative insights
Other jurisdictions, like Scotland and New Zealand, are proceeding with more caution in this context, in contrast to England and Wales, Australia and the US.Footnote 99
In New Zealand, some human rights protections applicable to AFR surveillance are less entrenched in the domestic legal framework than in England and Wales.Footnote 100 For example, individuals cannot use domestic human rights legislation to advance a judicial review of the effect of a piece of legislation or policy affecting their rights. Moreover, New Zealand's Privacy Act 2020 does not offer the same level of protection for the collection and processing of AFR data as the European Union's General Data Processing Regulation. The Privacy Act 2020 offers a single level of protection for all personal information without distinguishing AFR as involving the processing of sensitive biometric data, and does not require that the use of AFR must be ‘strictly necessary and subject to appropriate safeguards for the rights and freedoms of the data subject’.Footnote 101
Despite sharing a broad alignment with England and Wales in terms of legal structure – particularly as far as the applicability of human rights and data protection provisions to the police use of AFR is concerned – Scotland has imposed a moratorium on live AFR. While Police Scotland's 10-year strategy, Policing 2026, included a proposal to introduce AFR,Footnote 102 a Scottish parliamentary committee was highly critical of this plan. The Justice Sub-Committee on Policing found that the proposed live facial recognition software is known to discriminate against women and those from black, Asian and ethnic minority communities; that there is no justifiable basis for Police Scotland to invest in this technology; that prior to any decision to introduce it a robust and transparent assessment of its necessity and accuracy should be undertaken; that its potential impacts on people and communities should be understood, and overall that AFR would be a radical departure from the fundamental principle of policing by consent.Footnote 103 A subsequent response from Police Scotland indicated that, much like in New Zealand, the police were not using live facial recognition technology currently, and had no plans to do so. Police Scotland was also clear that it would ensure safeguards are in place prior to introducing AFR, and agreed that the impact of its use should be fully understood before it is introduced.Footnote 104
The experiences of New Zealand and Scotland serve to show that in England and Wales pervasive AFR surveillance is the product of both structural legal weakness and cultural permissiveness of surveillance. We endorse the approach of the Scottish Government that has emphasised the need for assessment of human rights impact prior to the introduction of any technology, not afterwards as occurred in England and Wales. Moreover, the parliamentary report's foregrounding of communities and consent to policing is key.
Such contrasting dynamics in terms of policing and politics within Great Britain are not unique. The situation in England and Wales can be seen to follow a pattern that emerged in respect of police use of DNA material, another form of biometrics, where likewise the onus was on the individual to challenge expansive laws and practices after the event. In contrast to AFR, however, the police were not pushing the boundaries of what was permitted in respect of DNA, or operating under any legal uncertainty, but rather were complying with the terms of what was deemed ultimately to be problematic legislation, in terms of human rights. So rather than a legal lacuna as regarding AFR, the law on DNA collection and retention was remarkably permissive.
In S and Marper v United Kingdom,Footnote 105 the ECtHR considered an Article 8(2) challenge to laws in England and Wales which permitted a non-intimate sampleFootnote 106 like a mouth swab to be taken without consent from a person who had been charged with, informed that they would be reported for, or convicted of a recordable offence.Footnote 107 Such DNA samples were kept indefinitely, regardless of the outcome of the investigation or subsequent criminal trial, due to their perceived value in crime control.Footnote 108 Two people, S (a child who was acquitted) and Marper (an adult against whom proceedings were not initiated), sought judicial review of the police decision to collect and retain DNA from them, based on the existing legal framework. Their claim was unsuccessful in the High Court, whose decision was upheld by the Court of Appeal, on the basis that the risks to the individual were not great and were outweighed by the benefits of retention. Similarly, the House of Lords dismissed their appeal, finding that there was no breach of Article 8 and that if such a breach had occurred it constituted just minor interference, and that retention was proportionate to its aims.Footnote 109 (This is akin to the Court of Appeal in Bridges.) In contrast, the Grand Chamber in Strasbourg held that such ‘blanket and indiscriminate’ retention of DNA violated Article 8.Footnote 110 Unlike in Bridges, the issue there was retention rather than collection and processing per se. Notably for present purposes, the European Court endorsed the Scottish approach, whereby non-conviction-based DNA retention occurred for serious suspected offences for a certain timeframe only.Footnote 111 DNA could be collected in Scotland from someone arrested and detained on suspicion of having committed an offence punishable by imprisonment, and such samples and information derived from them would be destroyed following a decision not to institute criminal proceedings or when proceedings did not end with conviction.Footnote 112 DNA retention was allowed after prosecutions which did not lead to conviction for certain sexual or violent offences only.Footnote 113
In Bridges, the Court of Appeal claimed that ‘the context of [S and Marper] is far removed from [Bridges]’.Footnote 114 While the context of the case itself might differ in respect of the particular sort of interference, the legislative and preceding context is comparable. So the more cautious approach of Scotland has been evidenced previously. While one could regard this as exemplifying the luxury of a smaller jurisdiction, with fewer issues of crime, this is not sustainable. Rather this embodies the impact of a more interventionist and rights-oriented Parliamentary culture, and less of a law and order discourse. The state of affairs in England and Wales demonstrates an abdication of responsibility by Westminster Government to legislate in this context, the reticence of the courts in respect of the public framing of surveillance, and the enduring agenda and institutional bias of police forces.
Conclusion
It is clear that the police enjoy a broad discretion in using AFR, as a consequence of the overlapping, implicit framework within which they operate. We argue that AFR should not be used in the absence of explicit authorisation by Parliament with full legislative process, and there needs to be ethical review prior to policing roll out.Footnote 115 This paper has identified a series of issues of law and political culture that have led to expansive deployments of AFR in England and Wales, without due regard for its varied human rights impacts. We do not seek to resolve the appropriate content of legislation specifically. But, when thinking of possible future regulation, we might draw on the way covert surveillance is regulated.Footnote 116 In brief, this could entail legal procedures at the national level as well as the local. In terms of the former, a national statutory basis for AFR and other police algorithmic use could provide democratic mandate, and an accompanying statutory code of practice or subordinate legislation for each type of algorithmic surveillance may provide specific rules, such as limits to specific targeted persons and locations; retention limits; and requirements that guide the necessity analysis undertaken by particular forces. At the local level, independent verification/authentication of the AFR or other algorithmic systems used could uncover and address potential accuracy and discrimination issues, with perhaps the development of an approved list of providers. Requirements for the independent approval and oversight of the proportionality and necessity of operation in accordance with the Code of Practice can also be incorporated into the statutory framework as appropriate, and could take the form of judicial authorisation, as in relation to search warrants.
The Divisional Court in Bridges opened its judgment with an astute observation: ‘The algorithms of the law must keep pace with new and emerging technologies’.Footnote 117 Beyond enticing the reader in, the line captures the significance of the challenge that faces lawmakers as new technologies increase the capacity of law enforcement agencies to subject citizens to scrutiny and control. Much like a good algorithm, English law should provide a clear process; a process for determining the limits of the police's uptake of new technology that is compatible with the rule of law and our existing international human rights obligations. In particular, with each new innovation in surveillance technology, English law should stipulate clearly whether or not it is permissible for police to deploy the technology and provide guidance on the circumstances in which it is appropriate to use the technology before it is operationalised. Though few technical innovations can be envisaged or predicted and so legislated for in advance, what can be done is the appropriate defining and controlling of police powers. As far as AFR is concerned, the ‘algorithms’ of English law suffer from numerous structural deficiencies and have failed to rise to the challenge of providing adequate regulation for expansive police surveillance.
