UK HR Two-Minute Monthly: April 2024

Case about allegedly racially biased AI settled

The claimant, who was black, worked as a food delivery driver from November 2019 to April 2021. The respondent required (and requires) its drivers to use its app to access work and to be paid. The app uses facial recognition software that requires users to take a real-time photograph of themselves, and then verify their identity, which is checked against the driver’s profile photograph stored by the respondent.

The claimant experienced difficulties with the verification process as, when comparing his real time photograph to the photograph stored by the respondent, the AI technology used by the platform frequently failed to recognise him. Following several failed attempts to use the facial recognition technology, the claimant was suspended from work. According to the AI technology, there were continuing mismatches between the pictures the claimant took to verify his identity, and his driver’s profile photo stored by the respondent. He lost his access to the app and was dismissed.

In October 2021, the claimant brought claims for indirect race discrimination, along with harassment and victimisation, also on grounds of race. The claimant alleged that the respondent’s policy of requiring employees to use facial recognition technology (that has inherent racial biases against people who are black, non-white or of African descent) put him and other black drivers at a disadvantage, as they are less likely to pass facial recognition tests for verification. His claims were backed by the App Drivers and Couriers Union (ADCU) and the Equality and Human Rights Commission (EHRC).

In July 2022, the respondent applied to have the claimant’s claims struck out at a preliminary hearing. It argued that the AI software was not the cause of the claimant losing access to the app, asserting that human error was responsible. The respondent argued this element of human error was unrelated to the claimant’s race. However, the strike-out application was unsuccessful.

A final hearing was listed for November 2024, but in late March the parties reached a settlement. Accordingly, the issue of whether the relevant facial recognition software was racially biased will not be tested, at least this time, at a tribunal.

Why this matters

This case highlights the challenges employers face in successfully implementing AI tools in the workplace. Whilst AI can have a positive impact on efficiency and output, employers should be cognisant that there are risks that AI technology can lead to discrimination and human rights abuses as flagged by the EHRC. Transparency with employees in the implementation and use of AI in the workplace is always advisable to guard against unlawful discrimination when AI is relied upon to manage employees.

Manjang -v- Uber Eats UK Limited

---

In whistleblowing dismissals does the decision maker need to know the substance/detail of the protected disclosure?

The claimant was employed as Vice President of Communications and PR by the respondent. On 14 August 2019, the claimant sent an email to two HR consultants engaged by the respondent making protected disclosures regarding the CEO. On 3 September 2019, one of the HR consultants informed the CEO that a complaint had been made about her management style by the claimant - the HR Consultant did not communicate the details/specifics of the complaint.

On 14 October 2019, the claimant’s employment was terminated by the CEO, ostensibly by reason of redundancy. The claimant brought a claim under section 103A of the Employment Rights Act 1996 (ERA) for automatic unfair dismissal by reason of whistleblowing, alleging that the true reason he was dismissed was that he had made a protected disclosure about the CEO, not that his role was redundant. He also alleged he had been subjected to detriments.

The tribunal held in favour of the respondent finding that, for the dismissal to constitute an automatically unfair dismissal or detriment due to whistleblowing, the CEO, who dismissed the claimant, needed to have a certain level of detail regarding the contents and substance of the protected disclosure. The tribunal found that, even though the complaint (the claimant’s email of 27 August 2019) was in fact a protected disclosure, the substance/detail of the disclosure was not communicated to the CEO, who was the decision maker. She was told there had been a complaint but the information she had was not sufficient to constitute a protected disclosure. It emerged that she only became aware there was a protected disclosure after legal proceedings had commenced.

Fundamentally, she could not dismiss the claimant for making a protected disclosure if she did not know he had made one, even though as a question of fact he had. There was no, or no sufficient, causal link between the two. The claimant, particularly under the test required by section 103A of the ERA, could not show that the sole or principal reason for his dismissal was making protected disclosures.

The tribunal did however find that there was no genuine redundancy, and that this was a pretext for a breakdown in the relationship between the claimant and the CEO.

The claimant appealed on various grounds, including that the tribunal had erred by requiring that the CEO should have received a detailed communication of the protected disclosure. The claimant contended that, as long as a protected disclosure had been made, this already qualified as a disclosure for legal purposes and did not need to “re-qualify” as one when it was passed to the decision-maker.

The EAT upheld the tribunal’s decision and dismissed the claimant’s appeal, agreeing with the tribunal that the fact the decision-maker was aware there had been a complaint was not enough to establish liability on the part of the respondent. It was found that the CEO (who was the decision-maker) was not sufficiently aware of the detail of the complaints made for the decision to dismiss to be “wholly or principally motivated by [a] protected disclosure”.

The EAT considered the relevant statutory wording from section 103A of the ERA:

“An employee who is dismissed shall be regarded as unfairly dismissed if the reason or, if more than one, the principal reason, for the dismissal is that the employee made a protected disclosure”.

The EAT interpreted this as meaning that the respondent needed to have at least some knowledge of the content of the disclosure to be liable for any whistleblowing dismissal or detriment.

However, the EAT emphasised that, where the decision-maker is deliberately kept in ignorance of the substance or content of the disclosure and a bogus reason for dismissal is invented, then the employment tribunal is permitted to go behind this and find, if it is the case, that whistleblowing is the true reason for the dismissal. In this case there was no such deliberate action on the part of the HR consultant or the respondent generally - it was simply the CEO’s decision to dismiss the claimant on very little information.

The EAT was keen to place safeguards against any potential loophole or gap, where a decision-maker might be misled or fed incorrect information. This would be a situation in which a protected disclosure had been made, but steps are taken so that the decision-maker is effectively kept in the dark and dismisses the employee for a different reason.

Why this matters

This case reinforces a key aspect of the legal framework for whistleblowing dismissal claims being that, for employers to be held liable for an automatically unfair whistleblowing dismissal, the decision-maker(s) must have knowledge of the substance of the disclosure made, not just of its existence. At the very least, the decision-maker(s) must have sufficient information to satisfy the legal tests for a protected disclosure.  

Having said this, employers should be wary of any attempt to manipulate matters by not informing the decision-maker about the specifics of the protected disclosure, as recent case law has shown that the tribunal can look behind any deceit or misinformation and still find the employer liable for a whistleblowing dismissal/detriments.

It is important to note that very recent case law has indicated that whistleblowing dismissals and detriments are treated differently in this area. The “looking behind the scenes” principle will be used in dismissal cases, but it is less likely to be used in detriment cases. We cover the case relating to this next month. 

Nicol -v- World Travel and Tourism Council 

---

“Continuing act” of discrimination must consist of acts that are discriminatory

The law relating to a “continuing act” of discrimination has been in existence since the original 1975 Sex Discrimination Act and now appears in section 123(3)(a) of the Equality Act 2010 (EqA). 

The broad idea behind a continuing act is that:

• Where there are several incidents of discriminatory conduct taking place over a period of time (such as repeated acts of sexual harassment) that conduct will be taken to have ended on the date of termination of employment – this goes to the issue of limitation periods, so that rather than just a discriminatory dismissal, earlier incidents that might otherwise be out of time can be captured and added as part of an overall claim. There can be more than a three-month gap between incidents constituting a continuing act;
• It also means, as alluded to above, that previous acts of discrimination can be taken into account, and the claimant can allege a pattern/campaign of harassment.  

Claimants pleading continuing acts is not unusual - as in this case there are sometimes a large number of acts/incidents which are alleged to represent a continuing act of discrimination. This means the limitation period starts to run from the date of termination, and the respondent is liable for multiple incidents of discrimination.

This case was concerned with several issues, but this article only focuses on continuing acts. 

The claimant claimed that she had been subjected to age and disability discrimination for a period of over twelve months and referred to various incidents as being part of an overall change management process at the respondent leading to her redundancy. The tribunal accepted that the incidents referred to by the claimant represented a continuing act of discrimination, noting that the acts had all been part of a change management/redundancy process. This point was appealed to the EAT.

The EAT’s concern was that each incident forming part of the continuing act had to be in itself a discriminatory act, and this could involve examining the motives of individuals. For example, if a grievance alleging discrimination were poorly handled and subject to delays, this might not be discriminatory. It would have to be shown by the claimant that the poor handling of a grievance and the delays were, in themselves, acts of discrimination, rather than acts caused by non-discriminatory incompetence or indifference on the part of the respondent. The individual or individuals managing the grievance might be doing so poorly, but is this motivated by discriminatory intentions? Is the poor handling of the grievance because of or related to the claimant’s protected characteristic(s)? If not, then that particular incident cannot be part of a continuing act. In short, just because a grievance alleges discrimination, it does not necessarily follow that poor and inadequate handling of that grievance is also discriminatory.

In reversing the tribunal’s decision, the EAT made three important points:

• There were different individuals at the respondent involved in the alleged continuing acts, and the individuals involved in the dismissal had no close connection to the individual responsible for another act of discrimination which started the continuing act. The continuing act also consisted of different types of discriminatory conduct (direct discrimination and harassment) and different protected characteristics. However, these factors in themselves would not preclude a continuing act. So there can still be a continuing act under section 123(3)(a) EqA where there are different protected characteristics, different types of discrimination, and different individuals involved;
• However, despite the above, it is necessary for the tribunal always to identify, by reference to the facts, the discriminatory nature of each act relied on. If there is no discriminatory element to the act (for example incompetent but non-discriminatory handling of a grievance) then that act cannot form part of a continuing act and the chain can be broken; and
• Following the 2002 case of Commissioner of the Police of the Metropolis -v- Hendricks, a continuing act does not necessarily have to be part of a discriminatory policy or “campaign” but it should ideally have some degree of underlying connection. In Hendricks, the connection was an ongoing state of affairs where female ethnic minority police officers were treated less favourably. In most cases involving one individual, this test is more likely to be satisfied because the connection will be the individual.

Why this matters

This case illustrates that a continuing act can be established even if different protected characteristics, different forms of discrimination, and different individuals are involved in the incidents making up the continuing act.

However, it is critical that the tribunal shows that each incident making up the continuing act is, in itself, a standalone discriminatory act, and tribunal must show each one to be so in its written judgment. It was this final point that cause the EAT to uphold the respondent’s appeal on this point, and to find that there was no continuing act.

Worcestershire Health and Care NHS Trust v Allen 

---

NEws roundup

Report on the impact of generative AI in the workplace

On 27 March, the Institute for Public Policy Research (IPPR) published a report on how generative AI (GenAI) might impact employment in the UK.

The report identifies two key stages in the adoption of GenAI - the current first wave, and a future second wave in which employers will integrate existing AI technologies further and more deeply into their processes.

Based on an analysis of 22,000 job tasks, the IPPR found that 11% of tasks had already been impacted, including routine cognitive tasks such as database management as well as more organisational/strategic tasks, such as scheduling management. However, the IPPR believes this could increase to 59% of tasks in the second wave, impacting non-routine cognitive tasks and higher-earning jobs.

In the “first wave”, the most likely to be affected are:

• Those with back-office, entry-level and part-time jobs, such as secretarial, administrative and customer service roles;
• Women, who work in a larger proportion of impacted roles;
• Young people, as employers may hire fewer people into entry-level jobs and introduce AI technologies instead; and
• Low to medium-level earners, as they are most likely to be replaced by AI.

With regard to the “second wave”, the IPPR has created three hypothetical scenarios. These range from the worst-case scenario, with full displacement, all jobs at risk, 7.9 million job losses and no GDP gains, to the best-case scenario of full augmentation, with all jobs at risk augmented to adapt to AI instead of being replaced, no job losses and an economic boost of 13% to GDP.

The IPPR’s modelling shows that there is no predetermined single path for how AI will impact the workplace and, in its report, the IPPR urges government action to avoid the worst-case scenario. It recommends that the government should develop a “job-centric” strategy to encourage job transitions and ensure that the benefits of AI automation are shared across the economy. This should include supporting green jobs and "social occupations", which are more difficult for AI to carry out, regulatory change and tax incentives/subsidies to encourage job augmentation rather than displacement.

Ringfencing some tasks might ensure continued human involvement, for example tasks involving ethical sensitivity and moral judgments, creative originality and artistic expression.

Awareness of changes to flexible working rules

A new survey from ACAS (28 March) has found that 70% of employees and 43% of employers are unaware that the law is changing to make it easier for employees to request flexible working at work.

The law is already in force and there is no longer a need for a qualifying period for employees to request flexible working. From 6 April the right now applies from the first day of employment.

There are other changes, including employees having the right to make two flexible working requests in a 12-month period, and the “response” times being reduced from 3 to 2 months.

There is a revised ACAS Code of Practice on requests for flexible working to support employers and employees through this change and other reforms.

Government proposes legislation to prohibit NDAs from preventing individuals reporting a crime

On 28 March, the government announced legislation that will be introduced "as soon as Parliamentary time allows" with the effect that NDAs cannot be legally enforced if they prevent the reporting of a crime. A disclosure will be permitted in any circumstances if it is relevant to criminal conduct, for the purpose of reporting a crime, or accessing support or advice. The legislation will ensure that information related to criminal conduct can be discussed with the following groups without legal action being taken:

• Police or other bodies that investigate crime;
• Qualified and regulated lawyers;
• Other confidential support services such as counsellors or medical professionals.

The legal effect of other parts of NDAs will not be affected.

The announcement appears to be a development leading on from the government's response to its consultation on NDAs published on 21 July 2019, in which it committed to legislate to provide that no confidentiality provision in an employment contract or settlement agreement can prevent an employee from making a disclosure to the police, regulated health and care professionals, or legal professionals.

Arguably this may have little effect on employment settlement agreements, as very few (if any) NDAs seek to prevent ex-employees from reporting/assisting in criminal matters, and an ex-employee is still free to make a protected disclosure, although the aim of this legislation is much wider.

However, areas covered in previous editions, where NDAs might be prohibited in cases involving harassment and/or bullying, could have a more profound effect on settlement agreements, but those proposals are yet to be clarified and seem a long way from legislation, although they might be closer to becoming a regulatory requirement in financial services organisations under FCA/PRA rules.

---

This article was written with Trainee Solicitor Jemima Rawding