Fish welfare: making farm controls auditable

The European Commission’s 2026 P23230 overview of farmed-fish welfare describes an uneven landscape: national requirements differ, dedicated controls remain uncommon in many countries, and welfare measures are difficult to compare. The report introduces no new legal obligation and is not an audit of France. Its practical question nevertheless travels well beyond the European Union: can a farm use the data it already collects to demonstrate animal condition, detect deterioration and justify corrective action? Mortality, behaviour, water quality and high-risk operations need to form an auditable decision chain rather than a disconnected set of records.

A European review, not a country ranking

The Commission’s Directorate-General for Health and Food Safety ran the project over 2023 and 2024. The team conducted six fact-finding missions in Norway, Germany, Greece, Denmark, Spain and Finland. It also analysed Commission databases, reviewed published material and consulted experts. The scope covered production, transport and slaughter, with particular attention to Atlantic salmon, rainbow trout, gilthead seabream, European seabass and common carp.

This design provides detailed examples but sets clear limits. The six visited countries cannot represent every European production system, and France was not visited. The findings therefore cannot be assigned directly to French farms or authorities. The report is most useful as a map of recurring governance gaps and workable approaches, not as a national league table.

At EU level, Council Directive 98/58/EC includes fish within the animals kept for farming purposes. The transport and killing regulations also contain general provisions that apply to fish, although the report highlights the limited detail available for many fish-specific situations. Operators must distinguish applicable law, national requirements, private standards and voluntary guidance. A recognised good practice does not automatically become a legal duty.

Fish-specific controls remain limited

Most systems examined did not yet conduct routine, dedicated fish-welfare checks on farms or at slaughter. Welfare was often considered during animal-health inspections or following complaints. Transport controls were more developed, but inspection tools frequently lacked defined values for variables such as oxygen, temperature or stocking density.

The absence of a common threshold does not make assessment impossible. It makes context essential: species and life stage, production system, biomass, flow, season, operation and cohort history. The same dissolved-oxygen value does not have the same meaning in a trout raceway, a carp pond and a marine cage. Likewise, low mortality cannot demonstrate good welfare if fish show injuries, abnormal distribution or sustained feeding disruption.

The review also identifies a skills gap. Inspectors and farm staff need to separate an alert, a plausible cause and a confirmed diagnosis. Without shared rules, two people may score the same fin erosion, swimming pattern or handling event differently. Training must cover what to observe and how to record it consistently.

Turning farm data into decisions

Many farms already record biomass, feed delivery, growth, mortality, oxygen, temperature and sometimes underwater video. The weakness identified by the Commission is not always missing data. It is the inconsistent use of those data to evaluate welfare. A measure becomes operationally useful when it is linked to a reference, a local investigation trigger, an accountable person and a predefined response.

An auditable welfare grid can combine four evidence groups. Animal-based evidence includes mortality and suspected causes, injury scores, body condition, appetite, activity and distribution. Environmental evidence covers oxygen, temperature, pH, nitrogen compounds, flow and technical incidents. Operational evidence follows high-risk events such as grading, pumping, vaccination, transfer, transport, pre-slaughter crowding, stunning and killing. Decision evidence records what was checked, which corrective action was taken, when the cohort was reassessed and what outcome followed.

A site should select a small number of repeatable indicators rather than adopt an unmanageable catalogue. Behavioural observations become more reliable when time, duration and scoring criteria are standardised. Technical alarms should be tested, and missing data should be visible rather than silently treated as normal. Corrective actions also need closure criteria: installing extra aeration or changing a handling procedure is only the beginning; the farm must verify whether animal and environmental measures improved.

This approach protects against a common analytical error. A change in swimming, feeding or mortality can trigger an investigation, but it does not identify a cause on its own. Water quality, infection, crowding, nutrition, equipment failure and recent handling can produce overlapping signals. The record should preserve competing explanations until evidence supports one of them.

What the EU good-practice code can — and cannot — do

The 2025 Code of good practices on fish welfare and fish welfare indicators complements the overview report. It organises general recommendations on feeding, handling, water quality, stocking density, mortality removal, transport, stunning and killing, and environmental enrichment. Its authors explicitly describe the guidance as non-species-specific and non-mandatory.

The code can therefore act as a completeness check across the production cycle. It cannot supply a universal threshold to copy into every farm procedure. Values and responses still need to reflect species, life stage, system and site risk. Where scientific evidence is incomplete, farms can define a cautious internal baseline, document why it was chosen and revise it as comparable observations accumulate.

Technology requires the same discipline. Cameras, sensors and automated feeding systems may reveal reduced appetite, unusual distribution or oxygen failure earlier than periodic manual checks. They do not replace clinical observation, causal investigation or a tested emergency response. An alarm without ownership and a response procedure is not a complete welfare safeguard.

A three-horizon action plan

Immediately, a farm can map high-risk stages, test alarms and backup systems, and select a short set of indicators that staff can score reliably. It should identify who reviews an alert and within what timeframe. The map should include stages that are less visible in routine production records, particularly loading, transport and slaughter.

Over the following months, teams can build site-specific baselines. Cohort comparisons and combined trends are more informative than isolated values. Investigation triggers should be framed as prompts to check the system, not as universal welfare limits. Training can use real farm examples, shared scoring criteria and incident-response exercises.

In the longer term, welfare monitoring should connect with fish health, biosecurity, maintenance, water-quality management, staff competence and continuous improvement. This integration preserves the context of an alert and prevents productivity data from being interpreted separately from animal condition.

For farms outside the EU, the exact legal references will differ, but the governance lesson remains useful: identify the enforceable baseline, separate it from voluntary standards, define evidence that is biologically relevant to the species, and show how that evidence changes a decision.

Limits and cautions

P23230 is a descriptive institutional review, not a controlled comparison or an exhaustive survey of all 27 EU Member States. Its conclusions draw partly on six missions and information gathered during 2023-2024. The report does not estimate the prevalence of each practice in every country and cannot determine the welfare performance of an individual farm.

Several advanced examples remain voluntary. Some monitoring technologies and stunning approaches are still being developed, and their effectiveness depends on species, fish size, equipment and applied parameters. Introducing them without local validation may move risk rather than reduce it. Finally, the current lack of harmonised thresholds never removes the duty to comply with applicable legislation, national rules, licensing conditions and site-specific requirements.

Conclusion

The Commission’s review does not simply call for more spreadsheets. It shows why fish-relevant indicators must be connected to verifiable decisions. A strong assessment describes animals, environment and operations; retains context; triggers an accountable response; and checks whether that response worked. Building this chain now can improve daily management while making farm practice easier to explain during an audit or official control.

How Vetofish can help

Vetofish can audit a farm’s welfare-monitoring system, help teams select indicators suited to their species and facilities, review handling, transport or slaughter procedures, and build an alert-and-corrective-action grid. The aim is to make decisions more consistent and traceable without presenting voluntary guidance as law or promising a guaranteed welfare outcome.