Industries / Food, farming and forestry / Food and beverage processing / Meatpacking
Slaughter
Slaughter is the work of stunning an animal so that it feels nothing, killing it by bleeding, and dressing the carcass: taking off the hide or feathers, the head and feet, and the internal organs, and splitting the body in two.
7
Milestones achieved, from 1860s to 2020.
2
To watch for: announced, not yet confirmed outside.
2
Open deliverables. Nobody has shipped them.
The road
AchievedWatch forOpen
Achieved
1860s
Hogs move on overhead rails
Cincinnati slaughterhouses moved hogs past a line of workers on an overhead rail, each worker doing one part of the job. A caption in the University of Virginia's American Studies hypertext dates this "line production and mechanical handling" to the 1860s, from an 1872 engraving, and notes an earlier picture of the overhead-rail conveyor in Harper's Weekly of 11 January 1868. In his 1922 book My Life and Work, Henry Ford wrote that his first moving assembly line, tried "along about April 1, 1913", came "in a general way from the overhead trolley that the Chicago packers use in dressing beef". It cut the time to assemble one flywheel magneto from about 20 minutes to 13 minutes 10 seconds, and later to 5.
Credit
Cincinnati pork packers and Chicago beef packers
1958
Stunning becomes US law
The Humane Methods of Slaughter Act, signed on 27 August 1958, declared it US policy that livestock be slaughtered and handled only by humane methods. It defined one such method as making the animal "insensible to pain" by a blow, gunshot, electric current, chemical or other rapid means "before being shackled, hoisted, thrown, cast, or cut". It also accepted religious slaughter in which the animal loses consciousness when a sharp blade cuts both carotid arteries at once. Federal meat inspection law now lets the Secretary of Agriculture refuse inspection to a plant that does not slaughter humanely.
Credit
US Congress (Public Law 85-765), amended by the Humane Methods of Slaughter Act of 1978
1971
Machines gut chickens
Evisceration, removing the internal organs, had always been done by hand in poultry plants. In 1971 Stork PMT launched its first automatic eviscerator, the P40. By the account of JBT Marel, which now carries that line, the fastest lines ran at about 4,000 birds an hour in the 1970s, 6,000 in the early 1980s and 8,000 by the end of that decade. A 2014 review in the journal Meat Science by Shai Barbut said automation had raised line speeds for beef, pork, sheep, poultry and fish, and put the fastest line he had observed at 13,500 broilers an hour.
Credit
Stork PMT, now part of JBT Marel
1999
Restaurant audits lift stunning
A captive bolt is a gun that drives a steel rod into the animal's skull to stun it. In a 1996 survey for the US Department of Agriculture, Temple Grandin found that only 4 of 11 beef plants (36%) made 95% or more of cattle insensible with one shot. Poor gun maintenance and awkward air-powered stunners were among the causes she found. From 1999 McDonald's, Wendy's and Burger King audited their suppliers' plants on a few scored measures, such as the share of animals stunned with one shot, moved with electric prods, or vocalizing (mooing or squealing). In 2008 audits of 32 US and Canadian beef plants, all 32 stunned at least 95% of cattle with one shot and 24 reached 99%.
Lessonby Grandin's account, stunning and handling "greatly improved" once buyers began scoring plants in 1999.
Credit
Temple Grandin of Colorado State University; the restaurant companies; the American Meat Institute, whose audit guidelines were used
2008
Undercover video forces recall
The Humane Society of the United States released undercover video of workers at a California slaughterhouse dragging, kicking and shocking sick and disabled cows to move them. The US Supreme Court later recorded that the video led to the largest beef recall in US history; federal rules forbid slaughtering any cattle that cannot walk (called non-ambulatory) for human food. England later made cameras compulsory: since 5 November 2018 every slaughterhouse there must run closed-circuit television (CCTV) in all areas where live animals are present and keep the footage for 90 days.
Lessonvideo from inside the plant became a tool of oversight, and in England a legal duty.
Credit
Humane Society of the United States; the UK government, for the English rule
2013
EU stunning rules apply
Council Regulation (EC) No 1099/2009, on protecting animals at the time of killing, applied across the European Union from 1 January 2013. Animals must be stunned before killing and stay unconscious until death, and plants must write operating procedures and check that stunning has worked. Each slaughterhouse killing more than 1,000 livestock units of mammals or 150,000 birds a year must name an animal welfare officer. Staff who handle, restrain, stun, shackle or bleed live animals need a certificate of competence. Slaughter without stunning for religious rites is allowed only as an exception.
Credit
Council of the European Union
2020
Court backs stunning for rites
A 2017 decree of Belgium's Flemish Region banned slaughter without stunning, including for religious slaughter, where it required reversible stunning that cannot kill the animal. Jewish and Muslim groups challenged it. On 17 December 2020 the Court of Justice of the European Union ruled that EU law lets member states require such stunning, as long as they respect the fundamental rights in the EU Charter.
Lessonthe court treated stunning before killing as the main animal-welfare aim of EU law and left each state to balance it against freedom of religion.
Credit
Court of Justice of the European Union, Grand Chamber, case C-336/19
Watch for
Argon in place of CO2
Most pigs are stunned in high concentrations of carbon dioxide (CO2), which pigs find aversive. The EU-commissioned PigStun project, with researchers, suppliers and processors from Germany, Denmark, Spain and the Netherlands, tested argon and helium against CO2. A review for Australian Pork Limited found that pigs reacted with less agitation to argon and helium but took longer to lose consciousness, and some still gasped. It judged argon the most practical gas for converting existing CO2 systems, noted that every alternative tested slowed throughput, and said PigStun's results were not yet peer reviewed.
Achieved whenargon stunning runs in a commercial plant at full line speed (the EU call set more than 800 pigs an hour) and welfare and meat-quality results appear in a peer-reviewed journal.
Claimed by
the PigStun consortium; reviewed by Ellen Jongman and colleagues (University of Melbourne and SunPork Group) for Australian Pork Limited, December 2025
AI cameras watch the pens
A project funded by the Australian Meat Processor Corporation (AMPC), an industry research body, with the company Impetus Animal Welfare, ran an AI monitoring system with CCTV and sensors at a commercial abattoir that kills calves, sheep and goats. AMPC says it gave continuous visibility of set welfare indicators and linked data from electric prods and stunning equipment to the footage, but that computer vision alone could not capture the full picture; it gives no accuracy figures. Video surveillance is becoming mandatory in stages under the Australian Animal Welfare Certification System. In Europe, Vion Food Group describes a similar tool, built with Deloitte, Eyes on Animals and the Dutch Society for the Protection of Animals, that scans all footage from unloading and the holding pens and flags clips for an animal welfare officer, also without published figures.
Achieved whena party outside the project publishes how often the system's flags match trained human reviewers for stunning and handling events, across more than one plant.
Claimed by
AMPC and Impetus Animal Welfare, April 2026
Open
01
Score bolt holes from photos
A camera station or phone app that photographs each cattle head or skull after slaughter, finds every captive-bolt hole, and measures how far each one sits from the recommended point on the forehead, where lines drawn from the top of each eye to the opposite horn cross. It gives a daily tally of shots by precision class and flags heads with more than one hole, a sign that the first shot failed.
The briefCredit · unclaimed02
Severity scores for handling records
An open, row-by-row table of US humane-handling enforcement records from the Food Safety and Inspection Service (FSIS), each incident coded by what went wrong (missed stun, animal hung while conscious, heavy electric prod use, no water, falls) and by severity on a fixed scale. A language model drafts the codes and two people check them. The table then tests consistency: whether incidents of the same severity led to the same action, a written noncompliance record or a suspension of slaughter, across plants and districts.
The briefCredit · unclaimed
Open 01 · the brief
Score bolt holes from photos
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
A camera station or phone app that photographs each cattle head or skull after slaughter, finds every captive-bolt hole, and measures how far each one sits from the recommended point on the forehead, where lines drawn from the top of each eye to the opposite horn cross. It gives a daily tally of shots by precision class and flags heads with more than one hole, a sign that the first shot failed.
Why it is still open
Fries and colleagues checked 8,879 cattle skulls by hand at two head-deboning plants; 64.7% and 65.3% had been shot in the ideal position and direction, and some skulls had more than one hole. They proposed checking shot holes on skulls after a day's slaughter as a measure of shooting precision. Of the 293 US humane-handling enforcement actions of 2018 to 2020, 188 concerned mechanical stunning, and where the cause of a failed stun was described, it was most often where the stunner was placed (Anderson, Kirk and Vogel, 2023). Grandin's guidance notes that the ideal point varies with breed and head shape and can be checked by splitting skulls. Searches of the web and GitHub for this file found studies that measure holes by hand and camera systems that watch stunning live, but no tool that measures holes from images.
Built with what exists
A phone or industrial camera with a printed scale marker in each frame, Ultralytics YOLO to find holes and landmarks (eyes, horn buds or the poll, the top of the head), OpenCV to measure distances, and a spreadsheet for the tally.
First step
With a plant's permission, photograph 100 heads beside a scale marker at the head-inspection table, have an assessor mark each hole and the ideal point, and train a first detector.
Done looks like
On 1,000 heads at one plant, the tool's count of holes per head matches a trained assessor's on at least 98% of heads, its distance from the ideal point is within 10 mm of the assessor's ruler measurement on 95%, and its precision class agrees with the assessor's with a Cohen's kappa (a measure of agreement beyond chance) of at least 0.8. Photos, labels and the tally are published.
Open 02 · the brief
Severity scores for handling records
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
An open, row-by-row table of US humane-handling enforcement records from the Food Safety and Inspection Service (FSIS), each incident coded by what went wrong (missed stun, animal hung while conscious, heavy electric prod use, no water, falls) and by severity on a fixed scale. A language model drafts the codes and two people check them. The table then tests consistency: whether incidents of the same severity led to the same action, a written noncompliance record or a suspension of slaughter, across plants and districts.
Why it is still open
The US Government Accountability Office (GAO) found in 2010 that inspectors had no objective tool, such as a numerical score, for judging plants and deciding on enforcement, and closed that recommendation as not implemented. FSIS's own follow-up reviewed 25% of its 2010 humane-handling records and found that 9% warranted stronger action than was taken. The Animal Welfare Institute has counted incidents plant by plant from public records (2016 to 2018, in a 2020 report). Anderson, Kirk and Vogel (2023) coded the 293 enforcement actions FSIS posted for 2018 to 2020 by species and cause, found the cause of half the stunning failures not clearly described, and did not grade severity or test whether like incidents drew like actions. No open, coded, row-level table with a measured coding accuracy turned up in searches for this file.
Built with what exists
Records posted on the USDA website and obtained under the Freedom of Information Act, pdftotext for extracting text, an open-weights language model fine-tuned on one GPU to draft codes (see the Note "One GPU to fine-tune"), and a spreadsheet for human checks.
First step
Gather one year of posted humane-handling records, write a coding rubric, and have two people code the same 100 records to test it.
Done looks like
At least 2,000 incidents coded; two human coders agree on cause and severity with a kappa of at least 0.8 on a 300-record sample; the model matches the human codes on at least 90% of that sample; and a published table shows, for each severity level and FSIS district, the share of incidents that led to a suspension.