Can You Drink the Tap Water in Brazil? Three Answers, Not One

Three parties answer this question and they do not answer it together. Federal law says what has to come out of the pipe. Each utility publishes what actually came out. What people here do with the glass has no statistics at all.

The standard in force, and the one everyone quotes instead

The rule is Annex XX of Consolidation Ordinance GM/MS No. 5 of 2017, rewritten twice since. The Ministry of Health states the chain itself: the standard in force is that annex "as updated by Ordinance GM/MS No. 888 of 2021 and by Ordinance GM/MS No. 2,472 of 2021".

Two details matter to anyone checking the text. Ordinance 888 was published on May 7, 2021 and republished on May 24 for having gone out "with an error in the original"; the versions differ in wording, not only layout, so the republication is the one to read. And the sampling annex printed on the 888 page was replaced that September — copying it means copying a dead table.

A third detail is about access. The Ministry's own link to the standard points to bvsms.saude.gov.br, and that host drops the connection during the TLS handshake, from a Brazilian line and a US server alike. The text here was read in the Diário Oficial, the government gazette, which carries both.

Ordinance 2,914/2011, still the number most English pages cite, has been superseded.

What the standard promises, and what the table allows

The definition settles half the question. Article 5 defines water for human consumption as potable water intended for "ingestion, food preparation and personal hygiene" — a drinking standard, not a washing standard.

Annex 1 separates two indicators that get mixed up constantly. E. coli has no tolerance anywhere: at the treatment plant outlet, in the network and at the points of consumption, the permitted value is the same absence in 100 mL. The annex calls it the indicator of fecal contamination.

Total coliforms work differently, and not the way a quick reading suggests. At the plant outlet they must also be absent in 100 mL, with no margin, as an indicator of treatment efficiency. The tolerance appears only downstream, where they indicate the network's integrity: for a system serving 20,000 people or more, absence in 95% of the month's samples; below that, one positive sample.

So "Brazilian law allows 5% of samples to be contaminated" is built by deleting half a table. Nor is the 95% a margin a company can work: article 27 says a negative retest does not cancel the original positive, and positives in a retest are not tolerated.

One requirement follows the water to you, and you can taste it. Article 32 obliges at least 0.2 mg/L of free residual chlorine "throughout the distribution system (reservoir and network) and at the points of consumption". The chlorine visitors notice is compliance, not a defect.

Where the water company's job ends

This is the part that makes the rest make sense. Article 5 defines the supply system as everything "from the catchment area to the building connections". The connection links the street network to the property's own plumbing, and that plumbing runs "to the user's water tank" — the user's, in the text of the standard.

The utility in Rio de Janeiro writes the same boundary in plain language, in the annual report it must publish: results for "the treated water delivered at the cavalete of your property". The cavalete is the meter box at the property line. Not the kitchen tap. Copasa, in Minas Gerais, describes its own reach the same way — a company describing itself, not a rule — from the catchment point to the water reaching its customers' properties.

Be careful what this establishes. The standard draws boundaries; it does not hand anyone the duty of maintaining a tank, and the words for "condominium" and "owner" never appear in it. It does reach partly inside the building, since sampling is required at points of consumption. The only direct federal command about a building's plumbing is article 26: it may not also be fed from another source.

That seam is the whole answer. The obligation is measured up to the cavalete. What happens between there and your glass — an old riser, a tank nobody has opened in years — is somebody else's, and unmeasured.

What actually came out of the tap, city by city

Copasa runs a public query by city and month: samples taken, how many fell outside the standard, and the average for each parameter.

Belo Horizonte, July 2026: 371 samples for total coliforms with one outside the standard, which is 99.73% compliance, and 371 for E. coli with none outside. Average free chlorine 1.26 mg/L, average turbidity 0.51 against a limit of 5. The parameter that slipped most was color — three of 373 samples. That single coliform result is not a breach: for a system that size the rule is 95%, and the page prints it under its own table.

Two small towns on the same query show what changes away from the capital, and it is not the compliance. Diamantina reported 44 samples that month and Tiradentes 12, both with nothing outside the standard. What scales with population is the sampling plan: the minimum required was 363 samples in Belo Horizonte, 35 in Diamantina and 10 in Tiradentes.

Rio de Janeiro publishes yearly instead. In the 2025 report for the city, analyzed equals compliant in all twelve months and all five parameters — turbidity, free chlorine, color, total coliforms and E. coli. March was 590 samples, April to September 270 a month, October and November 269.

Read the framing with the numbers. These are the utilities' own control samples, not the health surveillance's checks, and Águas do Rio serves 27 municipalities and 124 districts of the capital, not the whole city. Four cities say something about those cities.

How to check the city you are actually in

The question is answerable at city level because of a 2005 decree that obliges every utility to publish an annual quality report for each municipality it serves. Águas do Rio issues 27 a year, and its site carries 2021 through 2025.

Above the companies sit two federal systems that are easy to confuse. Sisagua, run by the Ministry of Health's Vigiagua program, is surveillance: the health sector checking, independently of the operator, with raw data in CSV, API, JSON and XML. SNIS, at the Ministry of Cities, measures the operator instead — IN084 is the incidence of coliform analyses outside the standard, and IN085 asks whether it collected the minimum number of samples. IN085 keeps IN084 honest, because an operator can post a flawless failure rate by testing rarely.

Open is not usable, though. The Ministry warns that some Sisagua files hold more rows than common spreadsheet software opens.

The layer with no statistics: what people here do

Behind this search sits a sentence you have almost certainly read: Brazilians don't drink tap water. It is in guidebooks, forum answers and travel blogs, often with a percentage attached.

We went looking for that percentage and it does not exist. IBGE, the federal statistics agency, publishes an aggregate catalogue of 9,298 tables, swept here in full. Crossing water against any term for the habit — drinking, filtering, boiling, bottled, potability — leaves exactly one table, about schools reporting whether they have drinking water. Crossing water against household leaves 109, every one measuring access: form of supply, sewage, bathroom. None asks what a person does before drinking.

There is no official basis to state it and none to deny it. A share of Brazilians who filter or boil is an estimate with no survey behind it, and this page will not add another.

So the honest version has three parts with three owners. The law requires drinkable water at the point of consumption. The utilities that publish results met it, in the cities and months checked here. The habit is unmeasured, and the last stretch belongs to the building.

Sources

• Ministério da Saúde, Vigiagua — the amendments in force. Water quality standard (accessed August 28, 2026)
• Diário Oficial da União — arts. 5, 26, 27, 32 and Annex 1, republication of May 24, 2021. Ordinance GM/MS 888/2021 (accessed August 28, 2026)
• Diário Oficial da União — the second amendment. Ordinance GM/MS 2,472/2021 (accessed August 28, 2026)
• Copasa — pick a city and a month. Water quality query (accessed August 28, 2026)
• Copasa — on its own reach. Water quality page (accessed August 28, 2026)
• Águas do Rio — the cavalete sentence and the monthly table. 2025 report, Rio de Janeiro (accessed August 28, 2026)
• Águas do Rio — one report per municipality. Report index (accessed August 28, 2026)
• Ministério da Saúde — formats and the size warning. Sisagua (accessed August 28, 2026)
• Ministério das Cidades — IN084 and IN085. SNIS glossary (accessed August 28, 2026)

Why this page cites a gazette rather than a ministry link is in how we research; it was written by Jeff Bruno in Curitiba. Daily life sits in Brazil basics, paying for it in money and payments — what a CPF is, the 10% on a bill. Then driving, phones, all guides.

What we could not verify

• Any figure for São Paulo. Sabesp's portal answers with the application shell: its pages for annual and monthly analyses carry no PDF links, and a headless browser did not finish loading them. The 2005 decree obliges it to publish and its index page exists — so this is the portal not serving the file, not the company failing to publish. The Northeast capitals were never searched.
• Who must clean a building's water tank, and how often. The federal standard draws the boundary and stops. The duty sits in state and municipal sanitary law, which we did not read, so we name no interval and no responsible party.
• The values behind the federal indicators. We read the SNIS formulas and confirmed the Sisagua data is open, but downloaded neither series. Every number above comes from two utilities.
• Whether Belo Horizonte's 1.26 mg/L of chlorine sits near the legal ceiling. Article 32 gives only the minimum, and we did not work through the chemical annexes that hold the maximum.