The great facial recognition experiment

Live facial recognition is a controversial technology that has been challenged in the UK courts and in Parliament. But the police are, once again, using football fans as a testing ground.

At the end of January, the House of Lords Justice and Home Affairs Committee (JHAC) wrote to Home Secretary James Cleverly expressing concern about the use of Live Facial Recognition (LFR) technology by police forces in England and Wales.

They highlighted three concerns in particular:

• “the absence of a foundation in law for the deployment of LFR”;

• “the lack of clear standards and regulation in respect of the use of LFR”;

• “the importance of consistent approaches to training in the use of LFR by police forces in England and Wales”.

Reporting the letter, the trade publication Computer Weekly referenced a previous investigation by the JHAC into the use of algorithmic technologies by police forces in England and Wales. The report summed up that investigation as revealing a situation “characterised by a lack of strategy, accountability and transparency from the top down”.

Those concerns and conclusions can only be fuelled by the response given to me and some former colleagues in the football supporters movement when we tried to obtain information on the deployment of LFR at football matches. It seems that, not for the first time, police forces are using football as a testing ground for techniques that can be more widely deployed.

In September 2023, I, along with other supporters of Tottenham Hotspur Football Club, received an email from the club ahead of the derby against Arsenal containing the following information.

“Dear Supporter,
 
Ahead of Sunday’s visit to the Emirates, we should like to make you aware that the Metropolitan Police Service will be using Live Facial Recognition technology at key locations in existing crime hotspots in Islington on the day of the match.
 
This technology will be used to find people who threaten or cause harm, including sex offenders and those who are wanted or have outstanding arrest warrants issued by the court. It will also be used to identify individuals who are subject to football banning orders. Signs will indicate where the technology is in use.
 
For more information on the Metropolitan Police Service’s use of Live Facial Recognition technology please visit their website here.”

For anyone who has taken more than a passing interest in issues around the policing and supervision of football fans, this communication set off alarm bells. If a measure is likely to be unpopular, chucking sex offenders and football hooligans in early to justify its use is a tried and tested method of disarming opposition. Because who in their right mind wants to stick up for sex offenders and football hooligans?

Quite why sex offenders were likely to be drawn in any number to attending a high-profile local derby was not explained. Note also the reference to fans on football banning orders – another tactic commonly deployed by the police to emphasise a threat to an audience that doesn’t understand the detail. FBOs can be issued for a range of offences, including being drunk while entering a stadium or the unauthorised sale of tickets, as well as violent conduct, disorder and indecent or racialist chanting. As of 1 August 2023, there were a total of 1,623 FBOs in place in the whole of England and Wales, according to official government statistics. Detailed datasets revealed that, as of the same date, 37 Tottenham Hotspur fans and 36 Arsenal fans were subject to FBOs.

But despite all of this, we were being asked to believe that a significant enough proportion of the 60,000 people attending a high-profile Premier League game could be sex offenders or one of the 73 fans of the competing clubs serving a banning order – a significant enough proportion to justify the use of highly controversial surveillance technology. As the punishment for breaking a FBO is severe, fans served with them don’t tend to go anywhere near football matches.

I was also aware of a case brought against South Wales Police by Cardiff resident Ed Bridges and backed by human rights organisation Liberty that has established the force’s use of LFR “breaches privacy rights, data protection laws and equality laws”. It was the world’s first legal challenge to the use of LFR, and the challenge was successful.

I wanted to find out what reasons the Metropolitan Police had for using LFR at the derby; what measures, if any, were being taken to address the issues highlighted in the Bridges case; and what the information gathered was telling them. Obviously the fine details of particular investigations could not be revealed, but this was a general deployment over a wide area involving tens of thousands of people and it could surely not be right that this controversial surveillance technique was being deployed with no real explanation or transparency.

I submitted a request under the Freedom of Information Act, and it was received on 11 September 2023. I asked the following questions:

1. Are there any plans to use facial recognition technology (FRT) at football matches?

2. If so, what matches?

3. Has FRT been used at any football matches?

4. If so, what matches specifically?

5. If FRT is to be used, are specific categories of supporter being targeted?

6. If FRT has been used, were specific categories of supporter targeted?

7. Has FRT been used, or is it intended to use it for, the prevention of specific crimes and if so which ones?

8. Has any hard evidence been collected about the effectiveness of FRT in any of the circumstances mentioned in this FOI request?

9. What, if any, measures have been taken or are planned to be taken to mitigate the effects of conscious or unconscious bias in the application of FRT and the processing of the information it gathers?

I received confirmation my request had been received, and that it would be answered within the statutory period of 20 working days.

On 9 October, 20 working days after my request was received, I received an email stating that the Met had been unable to complete its response. This was because it was considering an exemption on the grounds of law enforcement, and the response period would therefore be extended under the provisions of the act until 6 November 2023.

Then, on 11 October 2023, I received an email stating “due to the breadth of your request”, the Met would be “unable to proceed with your request as we require further information from you”. I was asked to clarify whether my request was focused on Overt Live Facial Recognition, a specific system in which “all persons passing the system’s camera(s) are analysed by the system with results being generated in tandem with events”. I was also asked if I was asking about use “inside or just outside the grounds”, and what I meant by “effectiveness”. Finally, the reply pointed out that the Act does not make provision for the expression of “context or opinion”, only “recorded information”.

I replied, saying: “Thank you for your reply to my FoI request. I’ve amended the questions to reflect your suggestions, and provided some explanation of why I have used the terminology I have.

“I do not know what specific type or types of FRT have been deployed at football matches. That is why I asked the question. If using the phrase ‘any type of facial recognition technology’ makes the question clearer, please use that definition.

“To clarify why I am asking about ‘effectiveness’, I assume there must be specific reasons for the deployment of such technology, both to justify its use and to enable officers to assess whether it has done the job they intended it to do. To be clear, this is not a request for context or opinion, but a question about whether recorded information shows that a decision has achieved the measurable outcome it set out to secure.

“I hope these explanations, and the questions amended and set out again below for clarity, enable you to proceed with processing my request.”

My questions now read:

1. Are there any specific plans to use facial recognition technology (FRT) in connection with Premier League, Championship, League 1 or League 2 fixtures, FA Cup, League Cup or European football matches played within London for the remainder of this current season. 

2. If so, what matches? 

3. Has FRT already been used in connection with PL, Championship League 1 or League 2, FA Cup, League Cup or European football matches played in London this or last season.

4. If so, what matches specifically? 

5. If FRT is to be used, are specific categories of supporter (ie, Risk Supporters) being targeted? 

6. If FRT has been used, were specific categories of supporter (ie, Risk Supporters) targeted? 

7. Has FRT been used, or is it intended to use it for, the prevention of specific crimes and if so which ones? 

8. Has any hard evidence been collected about the effectiveness of FRT in any of the circumstances mentioned in this FOI request? 

9. What, if any, measures have been taken or are planned to be taken to mitigate the effects of conscious or unconscious bias in the application of FRT and the processing of the information it gathers?

On 6 November I received an email telling me that the Met had been unable to complete its response, and that the deadline would be extended to 20 November 2023.

On 20 November I received an email telling me that the Met had been unable to complete its response, and that the deadline would be extended to 4 December 2023.

On 4 December I received an email telling me that the Met had been unable to complete its response, and that the deadline would be extended to 18 December 2023.

On 18 December I received an email telling me that the Met had been unable to complete its response, and that the deadline would be extended to 4 January 2024.

On 3 January 2024 I received an email telling me that the Met had been unable to complete its response, and that the deadline would be extended to 15 January 2024.

Then, on the 16 January 2024, I finally got my answers. Although when I say answers … I’ll let you read for yourself.

“DECISION

The MPS is required to apply the following exemption in respect of Question One (1) and Two (2);

·      Section 31(3) (Law enforcement)

The answers to Question Three (3), Four (4), Five (5), Six (6), Seven (7) and Nine (9) are already available in the public domain. This information is therefore exempt by virtue of Section 21 (Information available by other means). Further assistance can be found for these questions in the advice and assistance section below.

Information is not recorded that answers Question Eight (8). Further assistance can be found for this question in the advice and assistance section below.

As some data has been withheld as it is exempt from disclosure, this response serves as a Refusal Notice under Section 17 of the Freedom of Information Act 2000 (the Act). 

Please see the legal annex for further information on the exemptions applied in respect of your request.

Advice and Assistance

Question Three (3)

Has FRT already been used in connection with PL, Championship League 1 or League 2, FA Cup, League Cup or European football matches played in London this or last season.

The answer to Question Three (3) is already in the public domain. This information is therefore exempt by virtue of Section 21 (Information available by other means).

To assist you, the MPS have previously confirmed the answer to its use is that Overt LFR was used at Arsenal vs Tottenham on 24/9/23, as per the date listed on the Deployment Grid;

LFR Deployment Grid (met.police.uk)

Question Four (4)

If so, what matches specifically?

The answer to Question Four (4) is already in the public domain. This information is therefore exempt by virtue of Section 21 (Information available by other means).

To assist you, the MPS have previously confirmed the answer to its use is that Overt LFR was used at Arsenal vs Tottenham on 24/9/23, as per the date listed on the Deployment Grid;

LFR Deployment Grid (met.police.uk)

I refer you also to the Home Office link which confirms the use that day stating “At the Arsenal v Tottenham north London derby on 24 September 2023”:

Police use of Facial Recognition: Factsheet - Home Office in the media (blog.gov.uk)

This information is therefore exempt by virtue of Section 21 (Information available by other means).

Question Five (5)

If FRT has been used, were specific categories of supporter (ie, Risk Supporters) targeted? 

In terms of LFR used, LFR doesn’t ‘target supporters’ but instead it looks for people who are wanted for certain crime types that is driven by the intelligence case.

In the second page of the LFR Deployment Grid, there is a list of “LFR PURPOSE REFERENCES” on the 2^ndpage which provides the relevant categories that are held for this question;

LFR Deployment Grid (met.police.uk)

Rather than holding information in ‘specific categories of supporter’ as you have referred, the way the MPS holds information to assist with this question is to explain that where LFR was used, publicly available information is published within the ‘Purpose of Processing’ section of the DPIA:

Directorate of Legal Services (met.police.uk)

Additionally, you may can refer to the ‘Introduction, Aim and scope’ of the MPS LFR Policy Document to answer your request:

lfr-policy-document2.pdf (met.police.uk)

This information is therefore exempt by virtue of Section 21 (Information available by other means).

Question Six (6)

Has FRT been used, or is it intended to use it for, the prevention of specific crimes

In the second page of the LFR Deployment Grid, there is a list of “LFR PURPOSE REFERENCES” on the 2^ndpage which provides recorded comment that would assist with answering this question;

LFR Deployment Grid (met.police.uk)

This information is therefore exempt by virtue of Section 21 (Information available by other means).

Question Seven (7)

and if so which ones? 

General information is recorded and published (albeit not a specific list held) for this question and we advise you refer to the ‘Purpose of Processing’ section of the Data Protection Impact Assessment (DPIA) in regards to intention of use for the prevention of crime:

Directorate of Legal Services (met.police.uk)

This information is therefore exempt by virtue of Section 21 (Information available by other means).

Question Eight (8)

Has any hard evidence been collected about the effectiveness of FRT in any of the circumstances mentioned in this FOI request?

No recorded information is held to answer this question.

The MPS can advise that this is work that is ongoing and the term ‘effectiveness’ is very generic, wide and all encompassing. There is therefore no documentation the MPS can refer to that is recorded which answers the question in its format.

LFR is deployed as part of a wider policing operation and cannot be assessed on its own.

Question Nine (9)

What, if any, measures have been taken or are planned to be taken to mitigate the effects of conscious or unconscious bias in the application of FRT and the processing of the information it gathers?

This information is exempt by virtue of Section 21 (Information available by other means).

Recorded information pertinent to this request is within the published MPS Equality Impact Assessment (EIA) and Data Protection Impact Assessment (DPIA) which can be found in the link below (please navigate to the ‘Impact Assessments’ section);

Live Facial Recognition | Metropolitan Police

There is also the National Physical Laboratory (NPL) report on equitability of the LFR system; 

Operational Testing of Facial Recognition Technology (science.police.uk)

Thank you for your interest in the MPS.”

Had it really taken over four months and six missed deadlines to deliver this?

The answers assert that information on matters such as where LFR was used, against whom and for what purpose is already in the public domain, but does not confirm whether the information in the public domain is complete.

The response also says that LFR “looks for people who are wanted for certain crime types that is driven by the intelligence case”. Pretty vague stuff.

The most interesting answer is given to my question 8: “Has any hard evidence been collected about the effectiveness of FRT in any of the circumstances mentioned in this FOI request?” The response is that “No recorded information is held to answer this question.”

This was an ideal opportunity for the police to clearly state that ‘use of LFR was effective because … “. By not doing this they miss an opportunity to establish and emphasise a measurable outcome that could be used to justify the rollout of LFT to other areas, and to assuage public worries about use of the technology.

Whether or not the whole response is in the spirit of freedom of information is a subjective judgment, but I’d argue most reasonable people would recognise it’s not exactly falling over itself to provide direct and clear answers. But the admission that there is no recorded information to answer the question of whether or not LFR is effective is quite something. Attempting to obfuscate by asserting that “effectiveness” is “generic” doesn’t really cut it.

The answer to the question about whether any measures are being taken to counter bias in the application of LFR suggests that already published impact assessments have provided the answer. But the Bridges ruling and the JHAC letter show that serious questions still remain on this subject, and the answer here reinforces worries expressed in JHAC’s letter to the Home Secretary.

A second FOI request was submitted in the wake of the announcement that LFR was to be deployed at the North London Derby by my former colleague on the board of the Tottenham Hotspur Supporters Trust, Anthoulla Achilleos, assisted by Tristan Foot, a co-optee to the board.

Their questions were;

1. Please provide a copy of the Data Protection Impact Assessment for the use of Live Facial Recognition (LVR) technology at the Arsenal vs Tottenham Hotspur match on 24 September 2023.

2. If the LVR identifies a face from the "watchlist" but an officer determines that the LVR is incorrect, how is the image disposed of and how long are records held for?

3. What safeguards has the Authorising Officer (AO) put into place to protect the rights of those individuals whose personal data will be processed but are not on a "watchlist"?

4. Please describe in general terms the nature of the intelligence that led to a case being made for deployment of LVR at this particular moment.

5. What fair processing information will be made available in the location in which LVR is deployed and how will that be tailored to meet the diverse needs of a large cross-section of society, making reasonable adjustments where necessary?

6. Please provide a copy of the Equality Impact Assessment for the deployment of LVR in this case.

7. Following the recent cyber-attack on Manchester Police which resulted in large-scale loss of sensitive personal data, what additional measures has the Metropolitan Police taken to secure itself against loss of personal data?

8. What additional steps will be taken to protect sensitive personal data of children or disabled football fans during the use of LVR?

9. Has a particular risk of harm been identified at the time and place for this particular deployment of LVR?

10. If the answer to question 9 is yes, which policing methods other than the use of LVR were considered and why were they deemed insufficient?

11. Please confirm that officers will not penalise any individual choosing not to enter a zone of operation of LVR.

They received an answer within a month. It read:

“DECISION

Having considered the relevant information, I have decided to provide some of the information requested, however, part of the request has been refused by virtue of the following exemptions:

1.    Section 31(1)(a)(b) – Law Enforcement

2.    Section 21 - Information accessible to applicant by other means.

This response therefore serves as a Refusal Notice under Section 17 of the Act.  Please see the legal annex for further information on the exemptions applied in respect of your request.

REASONS FOR DECISION

Section 31 Exemption has been applied in relation to Q.4, 9 & 10.  This is because the MPS will not release details into the public domain which would give criminals an enhanced knowledge of our operational capabilities and highlight policing tactics as this would be detrimental to law enforcement. 

Section 21 Exemption has been utilised with reference to Q. 1, 2, 3, 6, 7 & 8. This is because the information requested is already available in the public domain and links have been provided below for your reference.

It is important to note that a Freedom of Information Act request is not a private transaction.  Both the request itself and any information disclosed are considered suitable for open publication, that is, once access to information is granted to one person under the Act, it is then considered public information and must be communicated to any individual should a request be received. Any information released under the Act is also published on the MPS website.

DISCLOSURE

With reference to Q.5, signage is placed outside the footprint of the deployment. QR codes within this allow for access to Facial Recognition documents that can then be viewed.  Leaflets are also available as well as through EIA (Equality Impact Assessment). Please see links below.

Links:

Data Protection Impact Assessment for LFR (Q.1)

Record keeping and Disposal (Q.2)

Data Protection & Policies & Processes in relation to Q.3, Q.8 & 11

Equality Impact Assessment in relation to Q.5 & 6

The appropriate System Security descriptions are detailed in this document, in relation to Q.7

Please see the section below titled ‘Legal Annex’ for further information on the exemptions applied in respect of your request.

This concludes your request for information and I would like to thank you for your interest in the MPS. 

Should you have any further enquiries concerning this matter, please contact foi@met.police.uk, quoting the reference number listed.”

It’s important to recognise that crime is becoming increasingly sophisticated, and that it is vital that crimefighting agencies utilise the best technology in the most effective way to ensure criminals don’t succeed. But what is also important is that the public, who crimefighting agencies are there to serve and protect, have confidence in the measures being used.

After the Bridges judgement, Liberty lawyer Megan Goulding said: “The Court has agreed that this dystopian surveillance tool violates our rights and threatens our liberties.”

And Liberty’s statement said: “In September 2019, the High Court found that South Wales Police’s use of facial recognition is not unlawful, but that facial recognition interferes with the privacy rights of everyone scanned by a camera, and 500,000 people may have been scanned (by May 2019) by South Wales Police.

“The Court found the current legal framework to be adequate, while warning that it would have to be subject to periodic review. Liberty challenged that ruling at the Court of Appeal in June 2020, arguing that it did not fully account for the ways in which the technology breaches our privacy and data protection rights and discriminates against people of colour.

Today, the Court of Appeal overturned the previous ruling, finding that the legal framework relied upon by South Wales Police does not protect our privacy rights.”

The statement also observed that: “The Metropolitan Police began regularly using facial recognition earlier this year, despite a review of its own trials finding the technology may be unlawful for similar reasons to those Liberty and Ed Bridges are raising.”

So there are serious questions about the legality of the deployment of LFR. That much was established by the Bridges case and by the concerns expressed by the JHAC. But it seems that, to add to the concerns, it is not possible to give a clear reason for its deployment or to provide any information on whether it is effective.

With little or no stated benefit from the use of LFR set out, it is legitimate to ask if the main purpose of doing so is not, in fact, to fight crime, but to test public resistance to its deployment.

• I’ve raised the issues set out here with a number of experts in the field of policing, surveillance and crowd control, and will be following up this edition of The Football Fan with more material.

Photo by Tim Mossholder on Unsplash