AI study raises fresh welfare concerns in UK greyhound racing
Bottom line
Greyhound racing data gets a sharper look with AI, and the picture is troubling. University of Melbourne researchers reported in a new Frontiers in Animal Science paper that AI agents helped assemble a population-level dataset covering 31,028 greyhounds and more than 1.26 million UK race starts from January 2022 through March 2026. Their analysis found that the Greyhound Board of Great Britain’s published fatality rate appeared flat when rounded to two decimal places, but rose 30% between 2022 and 2024 when examined at higher precision, alongside a 24% increase in absolute on-track dog fatalities. The study also found that 85.1% of dogs racing in the UK in 2025 were Irish-bred, median active racing duration before exit was 30 starts across 11.9 months, and most inactive dogs had no traceable public post-racing outcome. (frontiersin.org)
Why it matters: For veterinary professionals, the study is less about AI for its own sake than about what better data assembly can reveal in animal welfare oversight. The authors argue that national aggregate reporting can obscure track-level, trainer-level, and individual-dog risks, while the assembled dataset exposed track-specific adverse-event patterns and persistent attrition that had not changed since the UK industry’s 2022 welfare strategy began. That matters for veterinarians involved in racing oversight, welfare policy, sheltering, and rehoming, because incomplete visibility can limit surveillance, blunt prevention efforts, and make it harder to verify welfare outcomes after dogs leave the track. RSPCA Australia’s chief science officer, Dr. Suzie Fowler, said the method could help identify welfare concerns and assess whether industry or government changes are actually improving outcomes. (frontiersin.org)
What to watch: Watch for whether regulators respond with more granular public reporting, especially as bans or phase-outs move ahead in Wales and New Zealand and pressure builds on other jurisdictions to show clearer welfare outcomes. (gov.wales)
A new University of Melbourne study suggests AI may become an important tool for animal welfare accountability, not because it replaces expert judgment, but because it can assemble public data at a scale that exposes patterns regulators and industry summaries may miss. Published June 15, 2026, in Frontiers in Animal Science, the paper used AI agents under human oversight to compile a dataset spanning 31,028 licensed greyhounds, 1,267,122 race starts, and 51 months of UK racing activity. The headline finding was stark: while the Greyhound Board of Great Britain’s published fatality rate appeared stable at 0.03% from 2022 through 2024 when rounded to two decimal places, the underlying rate increased 30% when examined to three decimal places, with a 24% rise in the number of individual dogs killed on track. (frontiersin.org)
The work lands amid wider political and regulatory pressure on greyhound racing. The UK industry has pointed to its 2022 welfare strategy, “A Good Life for Every Greyhound,” and previously said 2022 data showed progress on injuries, euthanasia, and retirement outcomes. But the Melbourne researchers found no evidence that annual attrition changed after that strategy’s introduction. Outside England, the policy environment is also shifting: Wales announced in February 2025 that it would move to ban greyhound racing and introduced legislation in September 2025, while New Zealand announced in December 2024 that racing would be phased out, with cessation planned for July 31 or August 1, 2026, depending on the government source cited. (gbgb.org.uk)
The paper’s broader findings add context to the welfare debate. Researchers reported that 85.1% of dogs racing in the UK in 2025 were Irish-bred, and estimated that roughly 40% to 45% of non-coursing pups bred in the Republic of Ireland in 2021 and 2022 entered GBGB racing the following year. Median active racing duration before exit was 30 starts across 11.9 months, and most inactive dogs had no traceable, publicly documented post-racing outcome. The authors also said the GBGB holds unpublished data on retirement destinations, career-ending reasons, and individual-level injury and euthanasia figures, which they describe as a visibility gap with direct welfare implications. (frontiersin.org)
The study also argues that aggregation itself can become a welfare issue when it masks local variation. The authors said the governing body currently publishes welfare-relevant information as national totals, without meaningful public granularity at the individual-dog, per-track, or per-trainer level. Their assembled dataset, by contrast, identified track-specific adverse-event patterns and survivorship curves by cohort year. In practical terms, that means independent researchers could surface patterns that were technically public, but functionally inaccessible without large-scale data assembly. (frontiersin.org)
Reaction from welfare groups was swift and aligned with the paper’s conclusions. In the University of Melbourne release, co-author Dr. Simon Coghlan said reported fatalities are likely “only the tip of the iceberg” because public registries do not disclose post-race deaths linked to racing injuries, while Dr. Mia Cobb said the work shows that disclosure does not necessarily create real visibility. RSPCA Australia Chief Science Officer Dr. Suzie Fowler said the greyhound industry has long faced concerns over overbreeding, injuries, deaths, living conditions, and a lack of transparency after retirement, and that AI-based methods can help assess whether reforms are working. RSPCA UK’s Dr. Samantha Gaines said the findings reinforce calls to phase out the activity in England. (science.unimelb.edu.au)
Why it matters: For veterinary professionals, this story sits at the intersection of welfare science, surveillance, ethics, and data governance. The paper suggests that the quality of welfare oversight may depend not just on whether data exist, but on whether they can be independently assembled, audited, and interpreted at the level where harm occurs. That has implications for racetrack veterinarians, regulators, epidemiologists, shelter and rehoming partners, and clinicians who may see ex-racing dogs later in life. If post-racing outcomes remain opaque and welfare signals are reported only in broad aggregates, veterinarians may have a weaker evidence base for prevention, policy advice, and public communication. The study also offers a model that could be applied beyond racing, to other animal-reliant sectors where public-facing data are fragmented or selectively disclosed. (frontiersin.org)
What to watch: The next signals will be whether UK regulators or industry bodies expand public reporting beyond national summaries, whether 2025 UK fatality data deepen or contradict the 2022 to 2024 trend identified in the paper, and whether policymakers in other jurisdictions use AI-assisted evidence assembly to reassess greyhound racing oversight or phase-out timelines. (frontiersin.org)