WATER REUSE

Water Reuse in Food Production: The 2026 Codex Annexes and What a Water Scarce Site Can Defensibly Do

By Mthokozisi Nkosi, Food Safety Specialist & Lead Auditor, ASC Food Safety · 15 min read

CAC49, held in Geneva from 6 to 10 July 2026, adopted two new annexes to CXG 100-2023, one on fish and fishery products and one on fit for purpose assessment, safety management and recovery technologies. FAO had not republished the consolidated text when ASC checked on 10 September 2026. The Utilities, Water, Air, Steam and Lighting Risk Assessment Template (RA09) is R690.

What did Codex change about water reuse in 2026?

CAC49, the 49th Session of the Codex Alimentarius Commission, met in Geneva from 6 to 10 July 2026 and adopted two new annexes to CXG 100-2023, Guidelines for the Safe Use and Reuse of Water in Food Production and Processing. One annex covers fish and fishery products. The other covers fit for purpose assessment, safety management and technologies for recovering reusable water. Consequential amendments were made to the existing sections.

One point of fact before anything else, because it is wrong in most of what is circulating. The July 2026 session was CAC49, not CAC48. CAC48 met at FAO headquarters in Rome from 10 to 14 November 2025. If a supplier questionnaire or a customer specification lands on your desk citing a CAC48 water reuse decision, the reference is wrong and the person who wrote it has not read the source.

The second point matters more for what you can do this week. The adoption happened. The republished consolidated text of CXG 100-2023 carrying those annexes had not appeared on the FAO Codex site when ASC checked on 10 September 2026. That is a normal lag between a Commission decision and the publishing cycle, and it is not a reason to wait. It is a reason to be careful about what you quote.

The parent guideline is published, it is current, and like every Codex text it is a free download from FAO. Its fit for purpose principle is what actually decides whether your reuse scheme is defensible, and that principle did not change in July 2026. What the new annexes do is put more structure around the assessment and the recovery technology, and give the fish and fishery sector its own treatment. The full 2025 and 2026 change log sits in the ASC Codex and guidelines register, which carries 322 Codex texts and 60 other guidance documents, and the wider July 2026 picture is in our full list of every Codex change from CAC48 and CAC49.

What does fit for purpose water actually mean?

Fit for purpose is the heart of CXG 100-2023. Water used in food production does not have to be potable for every single use. It has to be fit for the specific purpose it is put to, judged against the hazards that purpose can transfer to food. The consequence is the part sites miss: the burden of proof sits on you, not on the water.

Most South African sites run the opposite logic. Potable water arrives, potable water is used everywhere, and the assessment is a certificate from the municipality. That is a single quality standard applied to every use in the factory, which is safe but expensive, and in a drought it is the reason a site stops production rather than reallocating water it could have used.

Fit for purpose replaces one blanket rule with a set of decisions. For each use you name the water, name the use, name the hazards that could transfer at that point, name the quality parameters that control those hazards, name the treatment that achieves them, and name the monitoring that proves it. Six named things. If any one of them is missing, the use is not defensible, whether the water is recovered or straight off the municipal main.

Fit for purpose is not a lower standard. It is a standard you have to argue for, one use at a time.

That is why reuse is a documentation problem before it is an engineering problem. I have seen good recovery plant sitting idle because nobody could write down why the recovered stream was acceptable for the use it was plumbed to, and I have seen unassessed reuse running quietly on sites that would swear they do not reuse water at all. The difference between the two sites is a piece of paper, and only one of them can defend itself.

Which process steps can accept a recovered stream, and which cannot

The decision turns on how close the water gets to food, and on what happens after it makes contact. Water that is removed by a later step is a different proposition to water that stays with the product. Water that becomes part of the food is the strictest case of all, because there is no subsequent step to fix it. Work through your uses in that order.

Take the four cases the fit for purpose argument turns on. Water used in an initial wash of raw unprocessed material is doing a soil removal job, and further washing, cooking, freezing or a controlled process step usually follows it. That combination gives a site the most room to argue for recovered water, provided the wash does not simply move contamination from one unit of raw material to the next.

Water used in a final rinse before packing is the last thing to touch the food. Nothing downstream corrects it, so the quality argument has to be made entirely on the water and its verification. Water that becomes an ingredient is stricter still, because it is not removed at all. It is bought, specified and approved the way any raw material is, with a supplier approval, a specification and analysis against that specification.

Water contacting a cleaned ready to eat surface is the fourth case and it behaves differently from the other three. The water never touches food directly, so it looks harmless on a flow diagram. In practice it is the last thing to touch a surface that ready to eat product will then touch, which makes it the finishing step of the clean rather than a utility. It is assessed with the hygiene rules for the zone it sits in.

Notice what the table is doing. It is not a permission list. It is a ranking of how much proof each use demands. A site can defend recovered water in an initial wash with modest treatment and clear monitoring, and the same site would need a much stronger argument, better treatment and far tighter verification to put recovered water anywhere near a final rinse. The general utilities picture, covering potable supply, ice, steam and compressed air, is in our utilities risk assessment guide for water, ice, steam and compressed air, and this article sits on top of it.

What the two new CXG 100-2023 annexes add

Two annexes were adopted at CAC49 in July 2026. The first covers fish and fishery products. The second covers fit for purpose assessment, safety management and technologies for recovering reusable water. Consequential amendments were made to the existing sections of the guideline. The consolidated text had not been republished when ASC checked on 10 September 2026.

Within those limits, the titles themselves tell you the direction. A sector annex for fish and fishery products means the Commission accepted that this sector needs its own treatment, which is unsurprising given how much water a fish plant moves and how much of it touches product directly. South African fish and seafood processors, particularly on the Western and Eastern Cape coasts, are the readers who should be watching hardest for the republished text.

The second annex title names three things a food site can act on immediately without waiting for the text: a fit for purpose assessment, a safety management approach for the reused water, and the technologies for recovering it. That is precisely the shape of the assessment set out in the rest of this article. Build those three now, keep them in editable form, and reconcile the numbering when FAO republishes.

Hazard identification for a recovered water stream

A recovered stream is not the same as the water that entered the factory. It has picked up soil, product residues, cleaning chemicals and organisms, and recirculation concentrates all of it. Hazard identification has to be done per stream, not once for the site, and it covers microbiological, chemical and physical hazards together, because the treatment that fixes one can create another.

Chemical carry over deserves its own paragraph because it is the hazard sites underestimate. Detergent and sanitiser residues that were harmless at the drain become a real question once the water is recovered, treated and sent back to a use where product contact occurs. Your cleaning chemical inventory, the rinse steps that follow each clean, and the point at which water enters the recovery loop have to be looked at as one system. That analysis belongs with the Cleaning and Disinfection Risk Assessment Template (RA03) at R690, which is where the chemical selection and the rinse justification already live.

The aerosol route is the one genuine hazard created by moving reuse away from product. Recovered water stored warm with long residence time is the condition Legionella needs, and wash down lances, yard hosing, dust suppression and cooling duty all aerosolise it. Where a recovered stream will be aerosolised, the assessment covers temperature control, tank turnover, disinfectant residual at the outlet and a Legionella monitoring regime. The exposure is to your own people rather than to the product.

Cross connection is the hazard that causes the incidents. A recovered water line and a potable line in the same plant room, with matching couplings and no labelling, will eventually be joined by somebody solving an urgent problem at three in the morning. Physical separation, distinct colour coding, non interchangeable fittings, backflow prevention and outlet identification are not paperwork controls. They are the ones that stop the event.

Choosing and specifying the treatment train

The treatment train is chosen after the hazards are identified, never before. Each stage is specified because a named hazard requires it, and the whole train is specified to deliver water at the quality the intended use demands. A train assembled from what a supplier had available, then justified afterwards, is the version that fails an audit and sometimes fails in production.

The stages usually run in the same order. Gross solids and organic load first, because everything downstream works better and lasts longer once the load drops. Then finer filtration sized to the physical hazards and to what the disinfection step needs to work. Then disinfection, chosen for the organisms identified and validated at the contact time and conditions the plant actually runs at. Then storage and distribution, which is where good treatment is most often undone.

Two pieces of chemistry set that order. Chlorine demand rises with organic load, so the residual falls away exactly when the water is dirtiest, and turbidity shields organisms from UV. The solids and organic load stage in front of the disinfection step is part of the disinfection specification, not a separate convenience.

Storage is the stage sites treat as neutral. It is not. A tank of treated recovered water, warm, with long residence time and no circulation, is a biofilm reactor with a lid. Tank turnover, level control, venting and vent filtration, drainability, cleanability, internal surface condition, dead legs in the ring main and the distance from the last disinfection point to the outlet all belong in the specification with the treatment stages.

Two other specification points are routinely missing. The first is what the train does with the chemicals used to clean the train itself, because recovery plant needs cleaning and that clean must not become the contamination event. The second is capacity under real conditions, meaning what the train delivers when the incoming load is at its worst, during the dirtiest product run of the week, and not what the sales brochure quoted at design load.

Verification monitoring for a recovered water stream, and how it differs from potable sampling

Monitoring for reused water works on two levels and sites regularly build only one. Operational monitoring runs continuously or at short intervals, tells the operator the train is working now, and drives an immediate action. Verification runs less often, uses laboratory methods, and answers whether the whole scheme is delivering what the assessment claimed. Both are needed, and neither replaces the other.

• Online or in line parameters read at operating frequency, with defined action limits and a named person who acts on them
• Laboratory verification at the point of use, not at the outlet of the treatment plant, because the distribution system is part of the risk
• Sample points designed into the system so the same point can be sampled repeatably, including the worst case point rather than the convenient one
• The incoming recovered stream sampled as well as the treated water, so you can see the load the train is actually being asked to remove
• Results trended over time rather than filed pass by pass, because drift is what tells you a membrane or a filter is ending its life
• A defined review of the whole scheme at a set interval and after any change to product, cleaning chemistry, volume or source
• Calibration records for every instrument the operational decision depends on

Set the limits before the first sample, not after the first surprising result. A reuse specification needs an operating target, an action limit and a reject limit for each parameter, with the action defined at each. Limits written retrospectively to accommodate results the plant happened to produce are visible to any experienced auditor, because the numbers sit suspiciously close to the data and the file carries no record of how they were derived.

The single most common weakness is verification sampling at the treatment plant outlet only. That result proves the train worked. It says nothing about the tank the water sat in for eleven hours or the ring main it travelled through afterwards. Sample where the water meets the product, and sample the point you would least like to sample.

What happens the day the treatment fails

Every reuse scheme fails eventually, and the assessment is judged on whether that was planned for. The failure procedure has to answer four questions in writing: how the failure is detected, what the water does the instant it is detected, what happens to product already made, and what has to be proved before the scheme restarts. If any of those four is unwritten, the scheme is unfinished.

Detection should not depend on a laboratory result that arrives two days later. The design intent is that the parameter which fails first is measured continuously, and that failure triggers a physical response rather than a note in a logbook. Automatic diversion of recovered water away from the food use, with a fallback to potable supply or a controlled stop, is the response an auditor expects to see designed in.

Failure also has a slow form that no alarm catches. A membrane fouling gradually, a disinfection dose drifting down, a filter approaching the end of its life and a storage tank whose turnover has quietly halved all degrade the water without ever crossing an instantaneous limit. That is why trending matters more here than in most monitoring programmes, and why the review interval belongs in the assessment rather than in somebody’s diary.

Product disposition is where the traceability system earns its keep. You need to be able to say which product was made with water from the affected period, which means the water system has to be traceable in the same way a raw material batch is. Sites that cannot answer this end up holding far more product than the failure justified, which is expensive, or releasing product they cannot defend, which is worse.

Does reused water raise the Listeria risk on a ready to eat line?

It can, and this is where a reuse project on a ready to eat site becomes a serious piece of work rather than a water saving initiative. Recovered water reaches floors, drains, equipment interiors and cleaned surfaces, which is exactly where Listeria monocytogenes establishes and persists. Water is also the vehicle that moves an organism from a harbourage site to a product contact surface.

CXG 61-2007, the Codex guidelines on controlling Listeria monocytogenes in foods, was fully revised at CAC49 and the published text carries REVISED 2026. The revision introduces a ready to eat definition built on what consumption is reasonably foreseeable. Codex publishes no list of the foods it reaches, so you apply the definition to your own range. Sites that never thought of themselves as ready to eat producers may now be inside that definition, and their water decisions travel with them.

The practical rule is straightforward. Any recovered stream that can reach a cleaned ready to eat surface, a drain in a high care area, or the air above open ready to eat product is assessed under the ready to eat hygiene rules and tied into the environmental monitoring programme. The sampling plan changes: you add sample points that follow the recovered water, and you test after cleaning, not only during production. The full argument is in our guide to Listeria after the 2026 Codex revision and the wider ready to eat definition.

The South African legal layer: water use, effluent and waste

A reuse project in South Africa sits inside a legal frame that has nothing to do with food safety schemes. Water abstraction and use, effluent quality and discharge, waste classification and the environmental authorisations that sit over all of it are governed by separate instruments under separate parent Acts, administered by different authorities, and the position depends on the site.

ASC does not publish a section number or a licence threshold in an article, because the instrument that applies to your site depends on your source, your volumes, your discharge route and what you already hold. Getting that wrong is not a food safety finding, it is a regulatory exposure with a different owner and a different set of consequences, and it is not something to resolve from a blog post.

What we do publish is the register. The ASC South African legislation register carries 413 instruments across 42 parent Acts and includes the water use, effluent, waste and environmental instruments alongside the food safety law. Use it to see the frame. Then ask us to confirm the position for your specific site in writing, so the answer is on a document with a date on it rather than in somebody’s recollection of a phone call.

Two other realities belong in the assessment even though they are not legal questions. Municipal supply interruptions and restrictions are ordinary operating conditions here, and borehole use is common. Both change the water safety argument, and both need to be in the same document as your reuse decision. Our South African food legislation guide maps the wider statutory picture.

What an FSSC 22000 or BRCGS auditor will ask for

Both schemes require water safety to be assessed and monitored, and neither will accept a reuse scheme that exists in the pipework but not in the management system. FSSC 22000 Version 7 was published in May 2026. BRCGS Global Standard Food Safety Issue 9, published on 1 August 2022, remains current, and Issue 10 has not been released. Under either, an undocumented reuse scheme is a finding waiting to be written.

• A current water schematic showing every source, every treatment point, every storage vessel and every outlet, with recovered and potable lines distinguishable
• The risk assessment behind the reuse decision, use by use, showing the hazards, the quality parameters and the reasoning
• The treatment specification, with each stage tied to a hazard it controls, and validation evidence for the disinfection step
• The monitoring and verification plan, the actual results, and the trend rather than a folder of certificates
• The failure procedure, with any real failure recorded, investigated and closed out
• Cross connection controls: separation, colour coding, labelling, backflow prevention and the inspection record for them
• Training records for the people who operate the treatment plant and for the people who use the water at the line
• The management of change record for when the scheme was installed or extended

The way the question arrives at audit is worth knowing. It rarely starts with water. It starts with an auditor following a pipe, asking where it goes, and finding that the answer is not on the schematic. Our complete guide to food hygiene audits covers how the visit is structured, and the 25 most common FSSC 22000 non-conformances shows how often prerequisite gaps like this one sit behind findings that look unrelated.

If your whole management system needs the underlying documents rather than one assessment, the toolkits carry them already indexed and cross referenced. The FSSC 22000 Food Manufacturing document templates toolkit is R6,350 with 260 or more documents and a clause mapped index, so a clause traces to a document and back. The BRCGS for Food Safety document templates toolkit is R6,550 with 220 or more documents. Each includes one hour of premium consultation with one of our consultants.

Auditors also ask who understands the scheme, not only who wrote it. Water and its recovery sit inside the prerequisite programmes, and two self paced courses cover that ground for a team that has never been taught it: Implementation of Good Manufacturing Practices at R1,950 and ISO22002-100:2025 The PRP Standard Requirements at R2,100. Both carry lifetime access and a QR verifiable certificate, and ASC charges no VAT on training.

A staged plan for a site that wants to start

Reuse projects fail when they start at the most attractive saving rather than the safest use. The order below starts where the risk is lowest and the learning is cheapest, and only moves closer to product once the site has real data from its own plant. Expect the first stage to take a few weeks and the whole sequence to run across a season.

1. Map and meterDraw the actual water system as it is, not as the drawings say. Meter the main uses. You cannot rank candidate streams without knowing volumes, and most sites discover uses nobody had accounted for.
2. Rank the candidate streamsScore each recoverable stream on volume, contamination load, ease of recovery and distance from product. The best first project is high volume, low load and far from food.
3. Write the fit for purpose assessmentFor the chosen stream and the chosen use, name the hazards, the quality parameters, the treatment and the monitoring. Record what Codex text you relied on and that the CAC49 annexes were pending at the date you wrote it.
4. Specify treatment against the hazardsBuild the train from the assessment, stage by stage, including storage and distribution. Get the validation evidence for the disinfection step from the supplier in writing before purchase, not after.
5. Engineer the separationPhysical separation, colour coding, non interchangeable fittings, backflow prevention and outlet labelling. Install these with the plant, never as a later addition.
6. Commission under verificationRun the scheme for a defined period with heavier sampling than the steady state plan, at the point of use and at the worst case point, and keep the results as the baseline.
7. Write and test the failure procedureDetection, automatic diversion, product disposition, root cause and restart criteria. Test it deliberately rather than waiting for the first real failure to test it for you.
8. Train the operators and the lineThe treatment plant operators and the people using the water both need to understand what the water is for and what it is not for. Record the training and assess whether it worked.
9. Review, then extendReview the data at the interval you set. Extend to a second stream only once the first has a full cycle of results behind it, and reopen the assessment when FAO republishes CXG 100-2023 with the new annexes.

Step three is the one clients most often ask us to do with them, because it is the step that has to survive an audit and a customer questionnaire. We will write it with your engineering and quality teams, or review one you have written. A full project is scoped and quoted as one figure rather than billed by the hour, and consultation starts from R480 an hour. ASC is rated 4.9 out of 5 from 350+ Google reviews, is SAATCA registered, is a FoodBev SETA Accredited Provider No. 587/00337/1900 and is B-BBEE Level 1 with 135% procurement recognition.

Read next from the ASC library

Frequently asked questions

Published by ASC Food Safety, South African food safety and quality consultants. This article is general guidance and not a substitute for certification-specific advice.