Industries / Finance and business services / Printing
Labels and packaging
Labels and packaging is printing on the materials that wrap and identify products: self-adhesive labels, folding cartons, corrugated boxes and flexible films such as snack bags.
7
Milestones achieved, from 1890 to 2025.
2
To watch for: announced, not yet confirmed outside.
2
Open deliverables. Nobody has shipped them.
The road
AchievedWatch forOpen
Achieved
1890
The first flexo press
Flexography prints from raised images on flexible plates, inked by a roller engraved with tiny cells (the anilox roller) that meters a thin, fast-drying liquid ink, which suits paper bags, film and board. Label Academy, a training program for the label trade, says the first flexo press was built in Liverpool in 1890 by Bibby Baron and Sons, and that later European presses, most of them German, used rubber plates and aniline dye inks, so the process was called aniline printing.
Credit
Bibby Baron and Sons; Grace's Guide records Bibby and Baron as a paper bag maker in Bury, Lancashire, that built its own machinery
1895
Gravure moves to cylinders
Gravure prints from an image engraved into a copper-plated cylinder. Karl Klíč worked out how to engrave such cylinders around 1879, and in 1895 he joined Samuel Fawcett and the Storey Brothers firm in Lancaster, England, to form the Rembrandt Intaglio Printing Company, which a 1913 trade journal says first offered the work in 1896. Gravure later set the quality mark for printing flexible packaging, one that flexo for a time could not match, by Bobst's account.
LessonThe firm kept its method secret rather than patenting it; dozens of rival gravure firms opened anyway, the Rembrandt company went bankrupt and was bought in 1932, so secrecy did not hold the lead.
Credit
Karl Klíč, Samuel Fawcett and the Rembrandt Intaglio Printing Company, by the Princeton University Library's Graphic Arts Collection
1952
Aniline printing becomes flexography
In the 1940s aniline dye inks were banned from US food packaging, and sales of aniline printing kept falling after safer colorants came in, by Label Academy's account. In March 1951 Franklin Moss, president of the plate maker Mosstype Corporation, asked readers of his company's magazine for a new name. More than 200 came in, a ballot chose among three, and "flexography" was announced on 22 October 1952 at the Packaging Institute's 14th forum in New York, by the account of Luminite, a plate-mounting supplier.
LessonA process can keep the bad name of an ingredient after the ingredient is gone; the trade changed both the inks and the name.
Credit
Franklin Moss and the Packaging Institute
1974
Photopolymer plates for flexo
Flexo plates had been made of molded rubber. In 1974 DuPont introduced its first Cyrel photopolymer plate, a light-sensitive plastic sheet that hardens where it is exposed, so plates could be made photographically. DuPont followed with plates imaged straight from digital files in the 1990s and plates processed with heat instead of solvents in 2000, by its own account in Labels & Labeling.
Credit
DuPont
2005
Carton ink reaches baby milk
A routine test by an Italian laboratory found traces of ITX (isopropylthioxanthone), a chemical used to cure UV inks with light, in Nestlé infant milk packed in Tetra Pak cartons. Nestlé recalled two million liters in Italy, France, Spain and Portugal, while Italian regulators began seizing about 30 million liters, by DairyReporter's account in November 2005. Tetra Pak said such UV inks had been used widely in the food industry for more than a decade, and switched to inks without such curing chemicals for products where migration was possible; a trade source told Food Manufacture the ink transferred when printed sheets were rolled up against the next layer.
LessonInk on the outside of a pack can still reach the food, so food packaging inks need their own rules; Germany adopted a list of substances allowed in printed food packaging in 2021, which applies from 1 January 2027.
dairyreporter.comfoodmanufacture.co.ukpackaginglaw.comgvw.com
Credit
Tetra Pak and Nestlé; found by Italian food control
2010
Recycled board carries ink oils
Maurus Biedermann and Koni Grob measured mineral oil in recycled paperboard used for food boxes. Unprinted recycled board held 300 to 1,000 mg/kg, mostly from the inks used to print newspapers, which held about 3,000 mg/kg, and in tightly packed boxes most of the lighter oil could move into the food within weeks. They derived a tolerable level in food of 0.6 mg/kg.
LessonRecycling carries the inks of one product into the next, so an ink that is harmless on a newspaper can end up in food.
Credit
Maurus Biedermann and Koni Grob, European Food Research and Technology, March 2010
2025
Digital labels, analog boxes
Smithers, a market research firm, counts about 8,100 narrow-web inkjet and toner presses printing labels and sleeves worldwide, giving digital 23.3% of the value of that work in 2025. Across all printed packaging, digital made 1.3% of output by volume and 4.1% by value, by Smithers' figures.
Credit
label printers and press makers; measured by Smithers
Watch for
Canon's inkjet press for boxes
Canon USA introduced the corrPRESS iB17, an inkjet press that prints corrugated board up to 1.7 m wide and 1 to 8 mm thick at 80 m per minute, with water-based inks and a primer for packs in indirect contact with food. It says the press will be on sale in the US in 2027.
Achieved whena box maker runs it in routine production and reports measured speed, uptime and waste.
Claimed by
Canon USA, October 2025
Hybrid presses take longer runs
A hybrid label press combines inkjet with conventional printing and finishing units in one line. Gallus, a Heidelberg subsidiary, says Artes Etichette in Italy installed the first Gallus Five in December 2025, runs it 15 hours a day at up to 100 m per minute, and regularly prints jobs of 10,000 to 60,000 linear meters; the printer's CEO says a complex job that took several hours to set up now starts in about 30 minutes.
Achieved whenthe printer or an outside party publishes measured setup times, waste and cost per job against its flexo presses.
Claimed by
Gallus, March 2026
Open
01
Open label defect benchmark
A public, labeled image set of printed label defects, such as missing ink, spots, streaks, color shifts, misregistration, die cuts out of place and waste matrix left on the roll, with clean reference images, and a baseline detector scored on how many defects it catches and how many false alarms it raises per 1,000 labels.
The briefCredit · unclaimed02
Table of ink migration results
A public, machine-readable table of measured amounts of printing-ink chemicals found in packaged food and in packaging, starting with photoinitiators such as ITX and with mineral oil from inks. Each row gives the substance, the food, the pack and print process, the amount measured, the method, any legal limit and the source table and page.
The briefCredit · unclaimed
Open 01 · the brief
Open label defect benchmark
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
A public, labeled image set of printed label defects, such as missing ink, spots, streaks, color shifts, misregistration, die cuts out of place and waste matrix left on the roll, with clean reference images, and a baseline detector scored on how many defects it catches and how many false alarms it raises per 1,000 labels.
Why it is still open
Label defect detection has been studied for years, for example on thermal-transfer labels in 2024, but searches of Hugging Face and Zenodo (November 2026) found no public set of label print defects. On GitHub the closest are two small single-product projects: a 2021 nameplate print check that compares each print with a reference image, and a 2025 inspection of laundry-pod labels.
Built with what exists
A machine-vision camera or recent phone over a hand-cranked rewind table; reject rolls from label printers, which their inspection systems have already flagged; labeling in CVAT or Label Studio; the open Anomalib library for anomaly detection and Ultralytics YOLO for a supervised baseline; a vision-language model that returns probabilities for yes-or-no questions, such as Gemini Decision 1, for a first pass at sorting images (see the Note "A thousand questions, one pass").
First step
Ask one label printer for a week of reject rolls flagged by its inspection camera and photograph 500 labels.
Done looks like
At least 10,000 labeled images from three or more label printers, eight or more defect types, released under an open license, with the baseline's catch rate and false alarms per 1,000 labels reported on a printer held out of training.
Open 02 · the brief
Table of ink migration results
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
A public, machine-readable table of measured amounts of printing-ink chemicals found in packaged food and in packaging, starting with photoinitiators such as ITX and with mineral oil from inks. Each row gives the substance, the food, the pack and print process, the amount measured, the method, any legal limit and the source table and page.
Why it is still open
The Food Packaging Forum's FCCmigex database maps which food contact chemicals have been detected in migration and extraction studies, covering literature up to May 2024, but its makers state that it holds no information on the concentrations measured. The European Food Safety Authority's 2023 opinion on mineral oil in food drew on 73,122 analytical results from 7,840 samples taken from 2011 to 2021, but reports them by food group, without the packaging or print process of each sample. Germany's list of allowed ink substances applies from 1 January 2027, and no public table puts published results next to it. A Europe PMC search for reviews of photoinitiator migration (November 2026) found reviews of test methods, not a shared table of results.
Built with what exists
FCCmigex's reference view to find studies; the Europe PMC and Crossref APIs; an AI model to draft rows from each paper's tables for a person to check (see the Note "Seven more flagged"); Python; a CSV file in a public GitHub repository. No lab work is needed.
First step
Export the FCCmigex references for ITX and for mineral oil hydrocarbons and enter the first 20 studies by hand to fix the columns.
Done looks like
At least 500 rows from at least 100 studies and official control reports, each traced to a table or page, with a 10% sample checked by a second person and the error rate published.