The Tap Question

On a single dated morning — 12 January 2026 — a tap in Dublin gained a number it had never had before. Under the recast EU Drinking Water Directive, water suppliers across the European Union became legally bound to meet, and to monitor against, a value written into Annex I: "Sum of PFAS — 0,10 µg/l."¹ An exceedance from that date is not a discussion. It is a breach of a parametric value, on a timetable the regulator sets.

A tap in Reading gained nothing that morning. The same 0.1 micrograms per litre — the same number, for the same family of "forever chemicals" — exists in England and Wales too. But it lives in a different kind of document. It is a guideline the Drinking Water Inspectorate publishes, a recommended action trigger, not a parametric value in the regulations.² Cross the Channel and the number is law. Stay in Reading and it is advice.

And the divergence is not only across the Channel. It is inside the United Kingdom. Scotland made the same number — "Sum of PFAS — 0.1 µg/l," at the consumer's tap — a statutory parameter in its own regulations, in force since 1 January 2023.³ So a household in Glasgow has carried a legally enforceable PFAS limit for three years; a household in Dublin gained one in January 2026; a household in Reading has neither. England and Wales sit behind both.

That is the whole of this report, and it is calmer than it sounds. The question a parent in Reading actually wants answered — is the PFAS in my tap water within a safe limit? — turns out, on the available evidence, to have a reassuring answer. The complication is that the reassurance and the gap are the same fact. To see why, you have to look at what a legal number does that an advisory number cannot, and who that difference hands the stopwatch to.

The same number, two kinds of document

Start with what is genuinely true, because the strongest case here is the case for the status quo, and it deserves to be stated plainly first.

England and Wales already use the EU's number. The Drinking Water Inspectorate — the DWI, the regulator for drinking-water quality in England and Wales — runs a tiered framework that classifies sites by their PFAS concentration: Tier 1 below 0.01 µg/L, Tier 2 below 0.1, and Tier 3 at or above 0.1 µg/L.² That 0.1 figure is the EU's figure. And the English panel is in one respect broader than Europe's: companies report a "sum of PFAS" based on 48 named compounds, against the EU's sum of 20.² More substances counted, same trigger number. There is also a genuine statutory backstop already in force — the general "wholesomeness" duty in the Water Supply (Water Quality) Regulations, which requires that water not contain any substance at a level that would constitute a potential danger to human health.⁴

So the number is applied. A wider panel is monitored. A statutory duty exists. A skeptic could stop here and say: the gap is a missing label on a number already in use.

The gap is narrower than a scare frame — but it is not nothing, and the difference lives in the kind of document the number sits in.

A statutory parametric value is a legal event. When the measured concentration crosses it, that crossing is a breach — automatically, on a date — and the response is not a cost-benefit judgement. The regulator controls the clock.

An advisory tier is a different instrument. The DWI's own guidance, issued in March 2025, describes the Tier 3 response as a "prioritised mitigation methodology to progressively reduce" PFAS through a "proactive and systematic risk reduction strategy."² That is the vocabulary of a duty discharged over time, on a schedule the company designs — "should consider," "progressively reduce," "prioritised." It is action, but it is action without an automatic breach on a fixed date. The general wholesomeness duty is real and statutory, but it is open-textured: "potential danger to human health" has no fixed PFAS value and no automatic trip point, so enforcing it on any given exceedance is a judgement, not a parametric event.

Same number. Two kinds of document. One hands the timetable to the regulator; the other leaves it with the party that would otherwise have to spend.

That distinction sounds abstract until you find it doing visible, filed work.

Where the missing number shows its weight

In its December 2024 final price determinations, the water regulator for England and Wales — Ofwat — formally created a route it named, in its own document, the "PFAS uncertainty mechanism."⁵ The mechanism lets water companies defer the recovery of PFAS treatment costs. Crucially, what triggers cost recovery is not a statutory standard. It is a future event: the determination provides for recovery "where investigations have been completed and the DWI has issued a new legal notice requiring works," with the reconciliation pushed "to the end of the control period."⁵

Read that trigger slowly, because it is the spine of the whole story. The obligation is routed through a future, discretionary legal notice — not a present parametric value. There is no fixed number that, once crossed, starts a clock the company cannot stop. The clock starts when, and if, the DWI issues a notice.

The mechanism did not appear from nowhere. Ofwat records that eleven named water companies proposed it.⁵ The fullest statement of the reasoning sits in a submission a ten-company group led by Anglian Water filed in August 2024, prepared by the engineering consultancy Jacobs. In its own filed words, the submission describes the DWI's PFAS rules as "non-statutory guidance" — guidance that "can be issued without consultation or other notice, and that water companies are nevertheless required to comply with," but which "results in massive uncertainty for water companies."⁶ The remedy the companies asked for was "an uncertainty mechanism to protect customers and themselves from the unforeseen costs that may arise from PFAS."⁶

It is worth being precise about what that filing does and does not say, because the temptation to over-read it is exactly the trap this story has to avoid. The submission does not say "we cannot commit to treatment because no statutory limit exists." It says three separate things: the DWI's rules are non-statutory guidance; that creates massive uncertainty for investment planning; and therefore a mechanism that lets the cost wait is the responsible thing to build. Joining those into a single causal sentence is my reading of the instrument, not a quotation from the consortium. And the consortium's stated purpose, on its own face, is protective — "to protect customers and themselves." When a regulated monopoly faces an obligation whose exact scope, trigger and timetable are not yet fixed even by the regulator, the financially careful thing — for the bill-payers who ultimately fund it — is not to lock in fixed treatment capital against an unknown specification. It is to recover efficient, actually-incurred costs as the requirement crystallises. That is what an uncertainty mechanism is. It is a rational response to a regulatory incentive structure; any monopoly facing an unpriced, undated obligation would ask for the same thing.

So the finding here is not about anyone's motive. It is about the instrument. An advisory regime permits deferral that a statutory value would foreclose, and here is the filed instrument that uses that permission. The missing statutory number is not an empty placeholder waiting to be filled. It is load-bearing. Its absence is precisely what routes a treatment obligation through a future discretionary notice rather than a present, dated breach.

And the body charged with oversight has said, on the record, that the decisive action isn't arriving. In April 2026, Parliament's Environmental Audit Committee — the EAC — published its assessment of the government's first national PFAS Plan. Its conclusion: "The Government's PFAS Plan disproportionately focuses on expanding PFAS monitoring rather than preventing or remediating contamination."⁷ The committee's chair put it more plainly: the plan is "short on decisive actions."⁷

Set the two documents side by side and the asymmetry is exact. On one side, a filed instrument that lets treatment cost wait until a future notice that the regime has no fixed date for issuing. On the other, the parliamentary committee whose job is oversight saying the prevention isn't happening. The missing number is doing work, and the body watching says the work it should compel isn't being done.

Why the statute is late, without anyone being blamed

There is a calmer reading of why England and Wales are here, and it is not neglect. It reads, on the record of how Britain has handled new drinking-water contaminants before, like a recurring pattern.

For new contaminants, Britain's drinking-water rules have tended to run guidance-first, with the enforceable value arriving later — often reactively, after a European directive, a World Health Organization revision, or a domestic incident moves first. Several of the binding standards already on the books for earlier contaminants — trihalomethanes, nitrate, lead — were imported rather than pioneered at home: pulled through by the EU framework, which itself drew on WHO guidelines, or tightened after an external instrument moved first. That is a reading of the historical sequence, offered as illustration of how the enforceable number has tended to arrive, not as a fixed law of British regulation.

On that reading, PFAS in 2026 sits on the same curve. The guidance already carries the right number; the statute is the part that waits — and in this pattern, the statute tends to wait for an external instrument or an incident to pull it through. The honest version is not "Britain never acts": when an incident forces it, domestic statute can follow promptly. It is that the enforceable value has tended to arrive reactively rather than being set ahead of the risk.

This is orientation, not alarm. It means the number a worried parent needs already exists, and already matches the EU's. What is missing is not the value. It is the legal status that would convert a deferrable liability into a present, dated obligation.

What is actually in the glass

Now the reassurance — and this is where the calm and the gap fuse into one fact.

In May 2026 a team at Imperial College London, led by Dr Alexandra Richardson, published what the Imperial team describes as the largest survey of PFAS in tap water across a UK city. It drew on nearly 8,000 measurements from 89 households across 28 London boroughs.⁸ Of 38 PFAS compounds tested, 11 were detected. Individual concentrations ran from about 0.6 to 6.0 ng/L; the total of all PFAS in any one sample ran from 3 to 35 ng/L — every result within the DWI's thresholds and well within the EU directive's value.⁸ "The amounts we're talking about are tiny," Richardson said.⁸

That is one city, 89 homes, 28 boroughs, a single study. It cannot be stretched to cover Reading, or rural Cambridgeshire, or all of England and Wales — and the absence of published per-area data for the rest of the country is itself part of the gap, not a footnote to the reassurance. But within its scope, the finding is genuine: London tap water is, on this evidence, comfortably within the number.

Before that "within the number" can mean anything, two scope questions sit under it, and both matter to a parent making squash.

First: the number is precautionary, not a harm cliff. The 0.1 µg/L value is technology-and-uniformity-driven — a single threshold set so the same number applies everywhere it can be reliably measured. It is a different kind of value from the European Food Safety Authority's health-based benchmark: a tolerable weekly intake of 4.4 nanograms per kilogram of body weight, for the sum of four PFAS, derived from a sensitive immune endpoint (a reduced antibody response to vaccination) rather than from a poisoning threshold.⁹ The two serve different purposes — one is a uniform regulatory line in the water, the other a health-based intake ceiling across all sources — and the parametric limit was not set as a direct conversion of the intake figure. "Within the limit" and "the limit is precautionary" are different statements. For adults, food — fish, eggs — dominates PFAS exposure; drinking water is typically the minor route.⁹ The one place that generalisation bends is young children: EFSA found toddlers and other children the most exposed group, the ages at which water plus diet together can approach the tolerable intake.⁹ That is the only part of this story that touches a baby bottle directly, and it dignifies the worry without inflating it — it is a reason to read your own number, not a reason to fear the tap.

Second: real exceedances exist, and the chain that catches them works. Near Duxford airfield in Cambridgeshire, a source drawn from permeable rock showed one type of PFAS — PFOS — at a level the public record puts as high as roughly 400 ng/L, the legacy of historic firefighting foam; a lower figure of about 131 ng/L also appears in the record for the same site, so the precise raw reading is best read as a range, not a single point. Either way it sits well above the DWI's Tier 3 figure of 100 ng/L. Cambridge Water said it reduced the level in residents' drinking water to around 10 ng/L by blending the contaminated source with cleaner ones.¹⁰ A real hotspot, found and brought back within the benchmark, without any statutory parametric value in force — the chain of monitoring and treatment did the work that the missing statute did not.

So here is the fusion. Put the reassurance next to the deferral mechanism and they stop being two facts in tension and become one observation. The honest answer to is my water within a safe limit? is, in London at least, yes — and that calm answer is precisely why the statutory backstop can keep waiting. A binding value would hand a dated, non-discretionary obligation to the regulator. An advisory one leaves the timing with the party that would otherwise spend. When the measured water is already within the number, the case for converting the guideline into a dated statute is exactly the case that can be deferred — there is no exceedance forcing the clock to start. The water being fine, in practice, is the reason the law isn't urgent. That is not a flaw in anyone's conduct. It is a property of the instrument: a precautionary number that is being met is the hardest kind of number to make mandatory, because nothing is breaching it. The reassurance is load-bearing too.

Why you can't easily answer your own question — and the filter sold into the gap

There is a second gap underneath the first, and it is the one a Reading household actually runs into. The EU directive carries a consumer-information duty: under Article 17 and Annex IV ("Information to the public"), residents are entitled to read their supply's water-quality results, online and updated at least once a year.¹ England and Wales have no equivalent requirement that puts a household's own PFAS figure in front of it against the benchmark. The data largely exists inside the monitoring system; what is missing is the duty to surface it per area, plainly, against a number you can read it against. The London survey is reassuring partly because it is rare — it is a research team publishing what the system holds but does not routinely show you.

Into that vacuum the filter market sells certainty, and the certainty is scoped in ways the box doesn't explain.

"Removes PFAS" is not one claim, because "PFAS" is not one number and "certified" is not one scope. Two standards govern the credible filter claims: NSF/ANSI 53, for drinking-water filters, and NSF/ANSI 58, for reverse-osmosis units. NSF tightened its certified total-PFAS reduction threshold in its 2022 NSF/ANSI 53/58 update, adding further PFAS compounds to the scope.¹¹ That is a real, meaningful tightening. But it remains scoped to specific named compounds at a specific limit — historically anchored to PFOA and PFOS, the long-chain compounds.

The chemistry decides what that scope misses. Granular activated carbon — the medium in most jug and under-sink filters — removes long-chain PFAS like PFOA and PFOS well, the US Environmental Protection Agency documents, "but shorter chain PFAS like Perfluorobutanesulfonic acid (PFBS) and Perfluorobutyrate (PFBA) do not adsorb as well."¹² The short-chain compounds break through the carbon bed early. Reverse osmosis, by contrast, is "typically more than 90 percent effective at removing a wide range of PFAS, including shorter chain PFAS."¹² So a filter can be truthfully "certified" against PFOA and PFOS while being weak on exactly the short-chain compounds that increasingly show up in the modern profile. "Certified" is not "removes everything you'd call PFAS." It is a claim with a named scope — and the named scope is what the advert leaves off.

What to do the next time this comes up

The instinct the filter advert wants is to buy first. The calmer move is to read first, because the number you'd read against already exists.

Find your water company's published water-quality data for your supply zone. Many water companies now publish per-zone PFAS monitoring; the figure to look for is the "sum of PFAS" result reported to the DWI, and any individual PFAS values. If your supplier doesn't publish a per-zone PFAS figure, you can request it. Read whatever you get against 0.1 µg/L — 100 ng/L — the number the EU and Scotland have made law. England and Wales carry the same number as guidance, so it is the right yardstick even though it is not yet statute where you live. If your zone's figure sits well below it — as the London survey's totals of 3 to 35 ng/L did — that is your answer, and it is the answer the filter advert was selling certainty against.

A filter earns its place where the local data shows a real reason: a documented exceedance, a known firefighting-foam or airfield catchment, a Tier 3 site near you. It is not a blanket purchase for a household already within the benchmark. For most London-equivalent supplies, the proportionate answer is no filter — read the number first.

If the data does warrant one, match the type to the problem and decode the claim before you pay (prices below are indicative, as of mid-2026):

• A certified jug or under-sink granular-activated-carbon unit carrying an NSF/ANSI 53 PFAS-reduction claim (roughly from £40) is the lighter option — strong on long-chain PFOA and PFOS, weaker on short-chain. Check the claim names the compounds, not just the word "PFAS."
• An under-sink reverse-osmosis unit carrying an NSF/ANSI 58 claim (roughly £330–£400) covers the wider range, including short-chain, at the cost of price and some wasted water at the reject ratio. It is the option where a confirmed short-chain profile or a documented hotspot warrants it.
• Where a product from a sustainability-focused range genuinely clears the same NSF/ANSI 53 or 58 scope, it sits alongside the others on the same test — and where it doesn't yet carry that certified scope, the honest position is that it isn't the answer for a short-chain problem until it does. The signal to trust is the certified compound list, not the brand.

And one honest limit on all of it: a filter relocates PFAS, it doesn't destroy them — the captured compounds end up in the spent cartridge. We found no widely available consumer filter currently certified against the precursor load that a more exhaustive lab method (the total-oxidizable-precursor assay) can reveal; if a certification expanded to cover that load, the scope gap on the box would narrow. The meaningful fix was never going to be 28 million households each buying and discarding their own filter.

The durable lever is the one piece of this that isn't a purchase. The number England and Wales need already exists, already matches the EU's, and is already being monitored against. What's missing is the legal status that would turn a deferrable liability into a present, dated obligation — and the per-tap transparency that would let a household read its own figure without waiting for a research team to publish it. That status would close on a specific date if the government's consultation produced a dated statutory standard; the absence of that date is the gap, confirmed on Parliament's own record. If a future national survey — not London-only — found scattered exceedances the calm anchor doesn't predict, the picture would tilt back toward the contamination frame; if the consultation delivered a dated standard, the enforceability gap would close the day it commenced. Neither has happened yet. The reader who wants the backstop has a real lever: the consultation is the document where the dated commitment will, or won't, appear.