Industries / Health / Veterinary
Farm animal care
Farm vets test herds for disease, prescribe the medicines given to animals that become food, and report outbreaks to the authorities.
7
Milestones achieved, from 1917 to 2022.
1
To watch for: announced, not yet confirmed outside.
2
Open deliverables. Nobody has shipped them.
The road
AchievedWatch forOpen
Achieved
1917
Every US herd tested
In 1917, after new tests made it possible to find infected cattle before they fell ill, the US Department of Agriculture began a campaign to wipe out bovine tuberculosis. Agents inspected nearly every cattle farm in the country and sent roughly 4 million reactors, cattle that reacted to the test, to slaughter without full compensation. The economic historians who studied it estimate that the benefits to farming alone were more than ten times the cost.
Credit
US Department of Agriculture; analysis by Alan Olmstead and Paul Rhode (Journal of Economic History, 2004)
1969
Feed drugs split from medicines
A UK committee chaired by Professor Michael Swann, set up after a drug-resistant Salmonella outbreak blamed on calf farming, reported in November 1969. Its main rule, still government policy in 1986, sorted farm antibiotics into "feed" drugs, of little use in treating people or animals, which could be sold without a prescription, and "therapeutic" drugs, which needed a vet's prescription.
LessonSwann set no limit on routine preventive dosing under a vet's prescription, and the campaign group Save Our Antibiotics says farmers moved from growth promoters to prescribed preventive use, so farm use kept rising.
Credit
the Joint Committee on the Use of Antibiotics in Animal Husbandry and Veterinary Medicine (the Swann committee)
2007
Q fever jumps from goats
Q fever is a bacterial disease that passes from farm animals to people. In the Netherlands it was diagnosed in dairy goats in 2005, and human cases began to climb in 2007. Q fever abortions were recorded on 30 dairy goat and dairy sheep farms from 2005 to 2009, and 3,523 human cases were notified from 2007 to 2009.
Lessonthe farm signal came two years before the human one, and the authors conclude that good monitoring and surveillance are needed to see the real size of such an outbreak.
Credit
review by Hendrik-Jan Roest and colleagues, Central Veterinary Institute of Wageningen UR (Epidemiology and Infection, 2011)
2010
Loans repaid for rural vets
A 2003 US law created the Veterinary Medicine Loan Repayment Program, which pays off the student loans of vets who work for three years where vets are scarce; a 2008 amendment gave priority to food animal practice. The first round, in fiscal year 2010, drew 260 applications and made 53 awards totaling $5,186,000; the program now pays up to $40,000 a year per vet.
Lessonan outside study of 2,163 food-animal shortage nominations from 2010 to 2021 found that only 19.3% to 36.8% were filled each year, and that areas with lower incomes, fewer farms and fewer cows were less likely to be filled.
Credit
US Congress; USDA National Institute of Food and Agriculture (NIFA)
2012
Dutch farm antibiotics halved
Antibiotic use to treat farm animals in the Netherlands doubled between 1990 and 2007. After drug-resistant bacteria such as MRSA were found in large numbers in farm animals, the government, vets' professional bodies and the livestock sector agreed compulsory and voluntary measures with set reduction goals. Use fell 56% between 2007 and 2012, against a target of 70% by 2015.
Credit
the Dutch government, veterinary organizations and livestock sector; account by David Speksnijder, Dik Mevius, Christianne Bruschke and Jaap Wagenaar (Zoonoses and Public Health, 2015)
2017
Vets sign off feed drugs
In January 2017 the US Food and Drug Administration (FDA) finished moving every medically important antimicrobial, meaning one also used in human medicine, given in feed or water from over-the-counter sale to a prescription or a Veterinary Feed Directive, an order only a vet can write. All growth claims came off the labels. FDA's sales report shows that sales of these drugs for food animals fell 33% from 2016 to 2017, and that over-the-counter sales fell from 8,000,326 kg to 271,280 kg.
Credit
FDA Center for Veterinary Medicine (Guidance for Industry #213 and the 2015 Veterinary Feed Directive rule)
2022
Europe bans routine dosing
From 28 January 2022, Regulation (EU) 2019/6 forbids using antimicrobials routinely, to make up for poor hygiene or farm management, or to speed growth. Antibiotics may be given to prevent disease only to an individual animal in exceptional cases. Metaphylaxis, treating a whole group when some animals are sick, is allowed only when the risk of spread is high and there is no other suitable option.
Credit
European Parliament and Council of the EU; European Medicines Agency
Watch for
Cattle TB vaccine and test
The UK's Animal and Plant Health Agency (APHA) is field-testing a BCG vaccine for cattle together with a new skin test, called DIVA, meant to tell vaccinated cattle from infected ones, which the current tuberculin test cannot do. Phase 1 ended in May 2022 and Phase 2 in November 2023, and APHA reports that the vaccine and tests were safe. Phase 3 covers 11 farms in England and Wales and at least 700 cattle, with on-farm work due to finish by summer 2026.
Achieved whenthe vaccine and DIVA test are authorized for use, and the test's accuracy in vaccinated herds is published and checked by a body outside APHA.
Claimed by
APHA for Defra and the Welsh and Scottish governments, March 2026
Open
01
Score AI on withdrawal times
A public test of AI chat models on drug withdrawal times, the number of days a farmer must wait after treating an animal before its milk, meat or eggs may be sold. Each question names a species, drug, dose and route taken from an approved US drug label, and the label's withdrawal time is the answer. The main score is how often a model gives a time shorter than the label, the mistake that puts drug residues into food.
The briefCredit · unclaimed02
Check AI flags reportable diseases
A set of written farm scenarios, each describing signs a farmer might type into a chat assistant. Some match diseases on the US National List of Reportable Animal Diseases, such as blisters on a cow's mouth and feet, a sign of foot-and-mouth disease; others match common, non-reportable problems. The score is whether the assistant tells the user to call a vet or the state veterinarian for the reportable cases, and how often it raises false alarms for the rest.
The briefCredit · unclaimed
Open 01 · the brief
Score AI on withdrawal times
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
A public test of AI chat models on drug withdrawal times, the number of days a farmer must wait after treating an animal before its milk, meat or eggs may be sold. Each question names a species, drug, dose and route taken from an approved US drug label, and the label's withdrawal time is the answer. The main score is how often a model gives a time shorter than the label, the mistake that puts drug residues into food.
Why it is still open
The USDA-funded FARAD program keeps the US Approved Animal Drugs Database and VetGRAM, an online tool for looking up established withdrawal times, so a trusted answer key exists. The published AI work here predicts how fast drugs leave the body from their chemistry (a 2026 cattle model reached a test R² of 0.69); it does not check the answers chat models give. Searches of PubMed, arXiv and the web found no such test; PetQA (2026), the largest new veterinary question set, covers only dogs and cats.
Built with what exists
Label data from FDA's Animal Drugs @ FDA database; VetGRAM for checking; current AI models through their APIs; Python to ask each question several times and grade the answers.
First step
Take the labels of 20 antibiotics approved for cattle, write 50 questions, and score two models.
Done looks like
500 label questions across cattle, pigs, sheep, goats and poultry, answered by at least five current models, with each model's share of correct answers and share of too-short answers published alongside the question set.
Open 02 · the brief
Check AI flags reportable diseases
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
A set of written farm scenarios, each describing signs a farmer might type into a chat assistant. Some match diseases on the US National List of Reportable Animal Diseases, such as blisters on a cow's mouth and feet, a sign of foot-and-mouth disease; others match common, non-reportable problems. The score is whether the assistant tells the user to call a vet or the state veterinarian for the reportable cases, and how often it raises false alarms for the rest.
Why it is still open
APHIS publishes the national list with case definitions and names state veterinarians as the contacts for reports. Farm assistants are being built, such as a 2025 goat-health assistant that scored 84.22% on its own test questions, but searches of PubMed, arXiv and the web found no test of whether such tools send a user with signs of a reportable disease to a vet. A November 2026 review in JAVMA found only 10 studies of generative AI in vet documentation and client communication, four of them on client communication.
Built with what exists
The APHIS list and its case definitions; general chat models and any farm assistant that offers trial access; Python to run each scenario several times and record the advice.
First step
Write 30 scenarios from the case definitions of ten listed diseases and run them on two public chat models.
Done looks like
300 scenarios, a third of them reportable, each checked by two vets; at least five assistants scored; the share of reportable cases sent to a vet and the false-alarm rate published with the scenarios.