Swiss animal testing hits record low as most stressful experiments rise

Switzerland used 507,863 animals in experiments in 2025, the lowest annual total since records began in 1983. The Federal Office for Food Safety and Veterinary Affairs released the figure on September 15, 2026, reporting a reduction of about 15,000 animals, or 3%, from the previous year.

The decline gives Swiss authorities a clear headline measure of falling laboratory animal use. Yet the same data shows a rise in the most severe category of experiments. Around 29,400 animals were involved in severity level 3 procedures, an increase of 7.4% in one year and the highest figure in 26 years.

Swiss law and official reporting classify experiments across four severity levels. Level 0 covers procedures that cause no distress. Level 1 involves mild distress, while level 2 covers moderate distress. Level 3 includes severe pain, prolonged suffering or serious impairment of an animal’s general condition.

The figures therefore track two separate developments in Swiss research. Fewer animals entered laboratories overall, while a larger number experienced procedures placed in the highest distress category. The national total does not show how many individual projects drove the increase, or whether the rise came from a small number of large studies.

Mice made up 66% of all animals used in Swiss experiments in 2025. Birds accounted for 13%, while rats and fish each represented 8%, according to the federal figures.

The composition of laboratory use matters because the overall decline was not shared evenly across species. The number of mice fell, while the number of fish rose significantly. The source does not provide the exact species-by-species totals, so the national reduction cannot be attributed to one broad change across all research programmes.

Mice remain the dominant laboratory species by a wide margin. Their use is closely linked to biomedical research, where scientists study disease mechanisms, treatments and potential therapies. Fish also appear in a range of research settings, and their increase contributed to the changing profile of Swiss animal experimentation even as the total number declined.

The published breakdown does not identify cantonal differences or provide a full account of the institutions behind each category. It does establish a national pattern: the fall in animal use has been driven mainly by a reduction in mice, while fish moved in the opposite direction. That distinction will matter in future assessments of whether Switzerland is reducing procedures broadly or shifting research activity between species.

About 73% of laboratory animals were used in research into human diseases. Cancer research and studies of neurological and mental illnesses were the largest fields identified in the federal summary.

Those priorities help explain why the severity figures deserve close attention. Research into serious diseases can involve procedures that produce severe pain, prolonged suffering or major impairment, placing them in severity level 3. The federal release does not specify which diseases, species or institutions accounted for the increase to approximately 29,400 animals.

The rise is substantial within the longer trend. Level 3 use increased 7.4% from the previous year and reached its highest point in 26 years, even while total animal use reached a record low. That combination suggests that counting animals alone gives an incomplete picture of the burden created by research.

Switzerland reports severity levels to distinguish the conditions animals experience during experiments. The classification provides a framework for public scrutiny, but the figures published here do not include project-level explanations, outcomes or details of measures used to limit suffering. More detailed information would be needed to determine whether the increase reflects new treatments, changes in study design, longer procedures or other factors within cancer and neurological research.

AI models are entering Switzerland’s debate over how to reduce live-animal use, but the 2025 figures do not establish how much they have changed laboratory practice. The Swissinfo report points to an AI mouse being developed by Empa, with the stated aim of reducing animal testing.

Such systems could help researchers model biological responses digitally, screen potential treatments or refine experiments before any live study begins. They may also support the replacement or reduction of animals in particular stages of research. The available figures, however, contain no measure of how many experiments were avoided because of AI, how many projects used computer models, or whether these tools affected the rise in severity level 3 procedures.

That gap is important for evaluating progress. A lower total may reflect improved study design, fewer animals per project or changes in research activity. It may also coexist with more intensive procedures involving the animals that remain. Without project-level data, the national statistics cannot show which explanation carries the most weight.

Empa’s work signals a Swiss investment in alternatives, while the record number of severe experiments shows that technological development has not yet removed the need for high-distress studies across biomedical research.

Switzerland’s animal testing policy will remain politically contested after the 2025 figures. A popular initiative submitted in 2024 calls for a comprehensive ban on animal testing. The initiative has not yet been scheduled for a vote, while the Federal Council recommends that voters reject it.

The latest statistics will give both sides material for the debate. Supporters of stricter limits can point to the increase in severity level 3 experiments, which reached approximately 29,400 animals and the highest level in 26 years. Those defending continued research can point to the overall total of 507,863 animals, a record low since official records began, and to the role of research into human diseases.

The next political discussion will also involve the credibility of alternatives. AI models and other non-animal methods may reduce live-animal use in selected stages, but the available 2025 data does not quantify their impact. Nor does it explain the project-level causes of the rise in severe procedures.

For Swiss voters, the eventual choice will turn on more than the headline total. The official record now contains two measures moving in different directions: fewer animals overall, and more animals exposed to the highest category of distress.