Bias, barriers & the potential for equality
AI Equality, Bias and AI Discrimination Case Tracker
This AI discrimination tracker examines both sides of AI’s relationship with equality: bias and exclusion, and the potential to prevent discrimination, improve accessibility and level the playing field.
Through legal cases and documented developments, I examine where AI creates or reinforces barriers, where it helps overcome them, and what the evidence actually establishes.
On this pageJump to a section
The question at the heart of this research
Does AI deepen inequality or help level the playing field?
The AI Equality, Bias and AI Discrimination Case Tracker brings both questions into view. Its purpose is to examine evidence of bias and exclusion alongside evidence of AI supporting fairness, accessibility and inclusion. The balance matters, and the answer may differ between systems, people and contexts.
Bias, barriers and discrimination
Cases about unequal treatment, exclusion, biased outputs or decision-making, and the safeguards surrounding AI and automated systems. Allegations and established findings remain clearly distinguished.
Equality, access and inclusion
Cases about AI’s use or proposed use to reduce human bias, prevent discrimination, remove barriers and support participation. Documented benefits, claims about benefits and requests for support remain clearly distinguished.
Weighing the evidence. The collection can illuminate benefits, harms and mixed outcomes in particular cases. Its case counts cannot establish whether AI advances or undermines equality overall.
Patterns in the public records
A wider view of the cases.
Explore the subject areas, contexts and jurisdictions represented in the public tracker. These charts describe this collection and should be read alongside its inclusion criteria and source limitations.
Area of law
Each record has one main subject category. These are editorial groupings, not a statement that different legal systems use the same rules.
Context of inclusion
Shows why each case is included: alleged harm or a challenge, a request for AI support, or a human decision in an AI setting. A support request is not proof of a benefit, and a challenge is not proof of harm.
Country
The jurisdiction of the recorded proceedings. These counts do not measure national rates of discrimination or the use of AI.
Counts, not conclusions. The charts cover the same public records as the table. They do not measure the prevalence of AI discrimination, the frequency of equality benefits, or the overall balance between them. They do not establish causation or rank countries and technologies by risk.
Equality, bias & accessibility
Explore the AI discrimination tracker.
Search the cases and follow the case-name links to the sources. The public table keeps four columns: date, case, country and brief details. Each summary identifies the relevant equality or bias issue and the position established by the sources, including allegations, findings and requests for AI assistance.
Dates identify documented events, not necessarily when proceedings began. An entry describes the stage established by its sources; later developments may change the position. The fuller research record explains the date and evidence basis.
| Date | Case | Country | Brief Details |
|---|---|---|---|
| 22-Jun-2020 | UK | "Paragraph 2.10 of the report explains:In early 2020, the Joint Council for the Welfare of Immigrants (JCWI), and the digital rights group Foxglove brought a legal challenge against the Home Office’s use of the Streaming Tool on the basis that:• it amounted to unlawful discrimination based on race contrary to the Equality Act 2010 (EA2010)• it contained ‘feedback loops’ which could drive further discriminatory decisions within the system.As part of its response to Judicial Review proceedings, the Home Office withdrew the use of the Streaming Tool across all entry clearance operations." | |
| 3-Oct-2018 | UK | The appeal concerned the lawfulness of the use of live automated facial recognitiontechnology (“AFR”) by the South Wales Police Force (“SWP”) in an ongoing trial usinga system called AFR Locate. AFR Locate involves the deployment of surveillancecameras to capture digital images of members of the public, which are then processedand compared with digital images of persons on a watchlist compiled by SWP for thepurpose of the deployment. | |
| 4-Sep-2020 | USA | The issue was whether false arrest was caused by an inaccurate facial recognition system. | |
| 1-Apr-2022 | USA | The issue was whether property management firm software and practices discriminated against renters receiving government housing subsidies. | |
| 1-May-2022 | USA | The issue was whether a company's hiring software unlawfully discriminated against older job applicants | |
| 1 June 2022 | US v Meta | USA | The issue was whether a company's advertising system discriminated against users in housing advertisements. |
| 1-Aug-2022 | FTC v Kochava | USA | The issue is whether a data broker sold sensitive geolocation data from millions of devices, potentially revealing private visits and exposing individuals to serious risks. |
| 1-Oct-2022 | USA | The issue is whether a bank unlawfully discriminated by race in home mortgage refinancing | |
| 1-Jul-2023 | USA | Issue: Did tenant screening company scoring cause racial discrimination. | |
| 1-Apr-2023 | USA | Issue: Does workplace applicant screening algorithm discriminate against African Americans, people over 40 and disabled individuals | |
| 25-Sep-2023 | USA | Issue: Did AI tools lead to housing discrimination. Did the tools reject internet rental applications from those receiving housing assistance. | |
| 2024 | UK | Issue was whether a Real Time ID check system in the UK which relied on Microsoft Real Time ID Check software and required drivers to upload a real time selfie when using the app which is checked against the driver’s profile photo caused indirect race discrimination, harassment and victimisation. | |
| 1-Jan-2025 | UK | Issue: whether Grammarly Pro was a reasonable adjustment. | |
| 1-Feb-2025 | Australia | Issue: whether paid version of ChatGPT (among other supports) met the statutory criteria of “reasonable and necessary” . | |
| 12-Jun-2023 | Netherlands | Issue: Whether Meta Ireland’s Facebook job‑ad delivery skewed by gender based on Global Witness tests | |
| 11-Sep-2023 | USA | Issue: whether a machine-learning algorithm subjected Black policyholders to additional administrative hurdles and delays in processing their claims. | |
| 23-Aug-2024 | Australia | Issue: court considered trans‑rights and equality and how gender‑identity discrimination law applies to AI‑mediated identity checks. |
This is an editorial research resource. It is neither an official register nor an exhaustive account of AI equality, bias and discrimination cases.
Different cases. Different questions.
Why the context matters.
AI equality and discrimination cases can concern a system’s operation, the safeguards surrounding its use, access to an assistive tool, or differences between human and automated decisions. The tracker keeps those distinctions visible without offering guidance on bringing or defending a claim.
A duty, not a bias finding
Bridges addressed the public sector equality duty around live facial recognition. The judgment expressly distinguished that issue from an allegation that the software itself was biased.
A request for AI support
In Nicholas, some disability claims succeeded, but the Grammarly Pro adjustment claim did not. The case brings proposed AI assistance into the equality debate; the requested adjustment was not upheld.
AI admission, human exclusion
In Tickle v Giggle, the AI initially admitted Ms Tickle. The appeal found direct discrimination in the subsequent human exclusion; special leave was refused in September 2026. This contrast does not establish that the AI system was generally fair.
A concise public view
The research behind the table.
I maintain a fuller research record with source references, procedural outcomes, technology descriptions, date explanations and quotation context. These fields are not all displayed in the public table.
Supporting research notes may be available on request, subject to availability and any applicable restrictions. The charts summarise the public selection. Their labels are editorial classifications, not legal conclusions.
Scope, sources & limits
How to read this tracker.
Does the tracker also cover AI advancing equality?
Yes. Its scope includes documented uses of AI to reduce bias, prevent discrimination, improve accessibility and advance equality, as well as alleged or established harm. It also includes claims about potential benefits and requests for AI assistance. Those are described according to the evidence; a hoped-for benefit or a refused support request is not recorded as a demonstrated success.
Can the charts show whether AI does more good than harm?
No. Weighing AI’s contribution to equality is a central research question, but this selected collection of legal records cannot measure the overall balance. Litigation is not a representative sample of everyday AI use. The charts organise the recorded cases; the sources and case context determine what can be said about benefits, barriers and mixed outcomes.
Does inclusion mean that AI discrimination was proved?
No. This collection includes allegations, settlements, procedural decisions, equality-body findings and requests for AI-related disability support. The brief details distinguish those positions. A case involving AI does not establish that AI caused discrimination, or that every person or organisation named acted unlawfully.
What technologies does this AI discrimination tracker cover?
AI, machine learning and related automated decision systems where the public source identifies a material equality, bias or accessibility issue. Some entries involve rules-based automation rather than machine learning. Others concern proposed assistive tools, or a human decision made in a service that also used AI.
How are cases counted?
One record normally follows one dispute through its procedural stages and appeals. Consolidated litigation is counted once in this collection. Separate claims by different people remain separate records, even when they concern the same organisation or technology. The charts count the public selection, not every research lead.
What does the date mean?
The public date identifies the documented event described in the research record. It may be a judgment, order, withdrawal or announcement. It is not necessarily the filing date. Where only an announcement is verified, its date is not presented as the date a settlement was signed.
Are the sources all judgments?
No. Sources include full decisions, filed documents, official regulatory material and identified party announcements. Reported entries are marked. For Johnstone, the relevant ChatGPT passage was available through published reporting, but the full official reasons were not independently verified in this review.
Why do allegations and findings appear together?
The purpose is to make relevant legal developments discoverable while preserving their different evidential weight. Settlements do not automatically establish liability. A claim surviving dismissal is not a merits finding. An equality body’s non-binding decision is identified as such.
What are the limits of the charts?
This is a selected research collection, not an exhaustive register or a representative sample. The charts cannot establish how often AI discriminates or advances equality, compare national risk, prove a trend, or calculate an overall balance of benefit and harm. Jurisdictions and legal grounds differ. “Harm alleged or challenged” describes why a record appears here, not a conclusion about causation.
How can I suggest an update or correction?
Please use the contact page with the case reference, a publicly available source, the relevant paragraph and the proposed correction. Please do not send confidential documents or material subject to reporting restrictions.
More research by Matthew Lee
Explore my other AI law trackers
International AI hallucinations
False citations and suspected or confirmed AI-related incidents across jurisdictions.
Judicial AI use
Reported uses of AI by judges, courts and judicial bodies.
UK AI hallucinations
Suspected and confirmed AI-related errors in UK courts and tribunals.
Government AI hallucinations
False or unreliable AI-related material in public-sector work.
Will AI replace judges?
Public views on AI, judicial roles and the future of decision-making.
Will AI replace lawyers?
Public views on lawyers, legal services and individual legal tasks.
Deepfake case law
Proceedings involving alleged or established synthetic images, audio and video.
Privilege & confidentiality
AI tools, confidential material and questions of legal privilege.
Prompt injection & hidden text
Hidden instructions and attempts to influence AI systems in legal contexts.
AI expert evidence
Cases and published guidance concerning AI-assisted expert material.
AI agents & liability
Autonomous transactions, authority and questions of responsibility.
AI & human rights
Cases concerning AI, automated decisions and human rights.
AI rights & legal personhood
Debates and legal developments concerning the status of AI systems.
AI chatbot harm
Claims involving alleged chatbot-related harm, safety and responsibility.
Alongside this equality, bias and discrimination tracker, these collections cover other subjects and jurisdictions. Each has its own scope. Entries may include allegations, suspected incidents, unresolved proceedings and public statements. Inclusion does not establish AI use, wrongdoing or liability.
Legal developments by area of law
The AI Law Reports
I also write The AI Law Reports, bringing together AI-related legal developments by subject. These ten registries provide routes into the relevant case material and legal reports and articles.
Employment law
Explore registry AI law registryCriminal law
Explore registry AI law registryHousing & property
Explore registry AI law registryPublic law
Explore registry AI law registryIntellectual property
Explore registry AI law registryTort & liability
Explore registry AI law registryFamily law
Explore registry AI law registryTax law
Explore registry AI law registryCivil litigation & procedure
Explore registry AI law registryImmigration & asylum
Explore registryEach card opens the relevant registry. Coverage varies by subject and develops over time; the registries are research and commentary resources, not official or exhaustive databases.
