New dyeing and finishing plants must comply from 1 September 2026; existing plants have until 2028


TCF POST Report


China now requires some textile plants to show that their treated wastewater doesn't kill fish embryos. The requirement is part of GB 4287-2026, the national textile wastewater discharge standard, which the Ministry of Ecology and Environment and the State Administration for Market Regulation published in June.

New, rebuilt and expanded facilities have had to comply since 1 September 2026. Existing plants must comply by 1 January 2028, or earlier where a provincial government decides so. The standard replaces four 2012 standards and extends coverage to chemical-fiber weaving and industrial textile products.

What the test involves

▪ Treated wastewater is diluted to one-sixth of the test solution.

▪ Newly fertilized zebrafish eggs are kept in it for 48 hours at 26°C.

▪ At least 90% of the embryos must survive.

▪ Plants under the strictest permit management must test at least twice a year, and all others at least once a year.

Who is covered

The limit applies to plants that discharge treated water directly into rivers, lakes or other water bodies. That means dyeing and finishing operations (excluding purely physical finishing) and shared treatment plants built specifically for the textile industry.

Mills that send wastewater to a shared textile-industry plant are not tested at their own outlet. The test is run at the shared plant instead. Mixed-industry treatment plants are not listed under the limit. For brands mapping their supply chains, this determines which party is responsible for the result.

Why regulators changed approach

Conventional limits measure one substance at a time, such as COD, ammonia or specific metals. They don't show how a full mixture behaves in a river. A whole-effluent toxicity test measures the combined biological effect, including from substances no one thought to look for.

The ministry said earlier textile standards had no overall toxicity control for dyeing and finishing wastewater, the higher-risk stream. The new test works alongside the chemical limits rather than replacing them.

The method isn't new. Germany swapped its adult-fish test for a fish egg test in 2004, and its textile rules still carry such a limit. The OECD adopted its Fish Embryo Acute Toxicity Test (Guideline 236) in 2013. What is new is that a major textile producer has made the test routine for permit compliance.

A failed test doesn't explain itself

A failed result shows that the effluent is acutely toxic, but not why. The cause could be a production chemical that survived treatment, a byproduct formed during production or treatment, or a flaw in how the treatment system is designed or run. Finding the source takes chemical input control, effective treatment, monitoring and root-cause analysis working together.

Certification bodies respond

Bluesign says it has required fish egg testing for more than a decade from chemical suppliers that discharge directly into water bodies. Its draft revised Criteria, open for public feedback until 30 October 2026, would extend an annual ecotoxicity requirement to:

▪ Directly discharging yarn and textile manufacturers (sites releasing more than 5 m³ of wastewater a day)

▪ Consumer product manufacturers with wet processing, including denim

▪ Testing through an accredited laboratory

▪ A fallback sequence for sites without access to fish egg testing: fish eggs first, then water fleas, algae and finally bacteria

The draft's fish egg limit matches China's one-sixth dilution threshold. It lists China's HJ 1069 method alongside OECD 236 and ISO 15088.

Questions brands should ask suppliers

▪ Does your wastewater go directly to a water body, to a textile-industry plant, or to a mixed-industry plant?

▪ Have you run a fish egg test with an accredited lab, and what were the results?

▪ What is your process for finding and fixing the cause of a failed test?

▪ Which chemical products drive your wastewater load, and were they assessed before use?

▪ Who reviews treatment performance, and how often? A single pass is a snapshot, not a system.