Why Does This Matter?

Every other case on this site is about AI deciding who gets hired. This one is about AI deciding who gets to keep working - a different, later moment in the employment relationship, with its own legal framework, its own harms, and its own reasons to matter.

This is not a resume-screening case - it's about identity verification for an existing worker

Mobley v. Workday and ACLU v. Aon are both about AI making decisions during hiring: which resumes get through, which applicants score well on a personality test or video interview. Manjang v. Uber Eats is different in kind, not just in degree. Manjang wasn't applying for a job - he'd already been working as an Uber Eats courier for over a year when the facial recognition system was introduced. The AI system in this case didn't decide whether to hire him; it decided, again and again, whether to let him keep proving he was still himself so he could keep working and getting paid. When the system and Uber's review process failed him, the alleged consequence wasn't a rejected application - it was the loss of his ability to earn an income at all, from one day to the next. This site treats it as a genuinely adjacent case worth tracking alongside the hiring-screening cases - both are about AI systems making or shaping decisions that affect who gets to work - but it would be inaccurate to describe them as the same type of case, and this page deliberately keeps that distinction plain rather than blurring it. (source: Employment Tribunal preliminary judgment, paragraph 9)

It's a real, funded, adjudicated example of the bias question outside the US

Every other case tracked on this site runs through United States law - Title VII, the ADA, the ADEA, or FTC/EEOC authority. Manjang's case ran entirely through the United Kingdom's Equality Act 2010, in front of an Employment Tribunal, a specialist UK court that handles workplace disputes. His claims (harassment related to race, victimisation, and indirect discrimination) are UK legal concepts with their own tests and case law, distinct from their nearest US equivalents. That makes this case useful as a real comparison point: the same underlying worry - that facial recognition and identity-verification technology may be measurably less accurate for people with darker skin tones - was tested through a completely different country's legal system, with different procedural tools (like the strike-out and deposit-order applications Uber used, twice, and lost, twice) and a different regulator (the EHRC) playing a role roughly analogous to the EEOC's in the US cases. (source: Employment Tribunal preliminary judgment, paragraph 8)

Uber tried twice to get the case thrown out before trial - and lost both times

Unlike Aon, which has seen very little public legal action beyond the initial filings, this case has a genuinely developed procedural record. Uber applied to strike out Manjang's claims (or at minimum force a deposit payment to keep pursuing them) at a preliminary hearing in May 2022 - and lost. Uber then appealed to the Employment Appeal Tribunal, which paused the appeal and sent Uber back to ask the original judge to reconsider; at that reconsideration hearing in September 2023, Uber tried again, this time backed by new evidence about its internal account-flagging systems - and lost again, on all but one narrow, technical point. Both times, the tribunal's core reasoning was the same: the reasons Uber gave Manjang for deactivating his account at the time didn't match the reasons Uber later gave the tribunal, and sorting out what actually happened would require a full trial with live evidence, not a decision made on the papers. That's a meaningful public record of a company's account being tested and found wanting well before any settlement - even though, importantly, it's still not a finding that discrimination occurred. (source: Employment Tribunal reconsideration judgment, paragraphs 78 and 94)

The case ended in a settlement, not a ruling on the merits

With a 17-day full trial scheduled for November 2024, Uber and Manjang settled the case in March 2024. That's an important, honest limit on what this case actually proves: a settlement resolves the dispute between the two parties, but it is not a court or tribunal finding that discrimination happened, and Uber did not admit liability. What the public record does show clearly is that the tribunal twice declined to say Manjang's case was too weak to go to trial - which is a real, if more limited, form of scrutiny than a settlement alone would suggest. See the timeline for exactly what was and wasn't decided at each stage.