IARC says glyphosate probably causes cancer. The EPA still says it probably doesn’t. A decade of litigation, a flawed regulatory process, and a trade agreement loophole later — New Zealanders are still caught in the middle.
In March 2015, the World Health Organization’s cancer research arm made a determination that sent shockwaves through the agricultural chemical industry. After a year-long review of publicly available scientific literature by a panel of independent international experts, the International Agency for Research on Cancer — IARC — classified glyphosate as “probably carcinogenic to humans.”
The United States Environmental Protection Agency still states glyphosate is “not likely to be carcinogenic to humans.”
Same substance. Same body of evidence. Two of the world’s most influential scientific and regulatory bodies. Opposite conclusions.
That contradiction sits at the heart of how glyphosate has been regulated — and how New Zealanders have been protected, or not — for the past decade. Understanding what those two phrases actually mean, how they came to diverge so sharply, and what has happened since, is essential context for anyone trying to make sense of the debate.
IARC’s “Probably Carcinogenic” Classification: What It Really Means
IARC’s classification system has five tiers. Group 1 means the evidence that a substance causes cancer in humans is established beyond reasonable doubt. Group 2A — where glyphosate sits — means the evidence is strong enough to conclude the substance probably can cause cancer, but falls short of the certainty required for Group 1. Group 2B means it possibly can. Group 3 means it cannot be classified. Group 4 — applied to only one substance in IARC’s entire history — means it probably cannot.
To be clear about what IARC is and is not measuring: it assesses hazard, not risk. Hazard is whether a substance has the capacity to cause harm. Risk is the probability of harm at a given level of exposure. IARC determined that glyphosate has the capacity to cause cancer — specifically, that there is limited evidence in humans and sufficient evidence in experimental animals, supported by strong evidence of genotoxicity, meaning glyphosate can damage DNA.
IARC also uses only publicly available, independently peer-reviewed studies. Industry-funded research that has not been published and made available for independent scrutiny is excluded by design. That transparency is a feature, not a flaw — it means the classification can be independently verified by anyone with access to the scientific literature.
The EPA’s “Not Likely” Finding — and Why a Federal Court Rejected It
The EPA’s conclusion sounds reassuring. But the story of how it arrived there is considerably less so.
The EPA’s 2020 assessment — which concluded glyphosate presents “no risks of concern to human health when used in accordance with its current label” — relied heavily on a body of research that included a significant volume of unpublished industry studies submitted directly by Monsanto and other manufacturers. Unlike IARC, the EPA’s review incorporated data that the public and independent scientists could not access or verify.
In June 2022, a three-judge panel of the US Ninth Circuit Court of Appeals ruled that the EPA had failed to follow its own established guidelines for determining cancer risk, had ignored important studies, and had discounted expert advice from its own scientific advisory panel. The court found the agency had applied “inconsistent reasoning” in its assessment. The EPA subsequently withdrew its interim decision and was ordered to start its review again.
That review is ongoing. The EPA’s website continues to state that glyphosate is “not likely to be carcinogenic to humans” — language shaped, as the court found, by a flawed process. It has also since emerged that a key paper central to the EPA’s cancer classification — relied upon to argue glyphosate posed no cancer risk in animal studies — was ghostwritten with significant involvement from Monsanto. The paper was formally retracted.
This is not a minor procedural footnote. It means the regulatory conclusion that has underpinned the world’s permissive approach to glyphosate was built, in part, on a foundation the courts have since found to be scientifically inadequate.
The Roundup Cancer Lawsuits: US$18 Billion and No Admission of Liability
Between these two regulatory positions sits one of the largest chemical liability cases in history.
Since 2018, Bayer — which acquired Monsanto that year — has faced more than 170,000 lawsuits in the United States from people who developed non-Hodgkin lymphoma after using Roundup. By 2020, Bayer had paid approximately US$11 billion to settle the majority of those claims. Jury verdicts in cases that went to trial have totalled billions more, with a California appeals court as recently as November 2025 affirming a US$28 million verdict for a man who developed non-Hodgkin lymphoma after decades of Roundup use. In February 2026, Bayer announced a further proposed settlement of US$7.25 billion to resolve current and future claims.
In every settlement, Bayer has admitted no liability. In law, that matters — settlement is not an admission that the product caused cancer.
But the scale of what has occurred invites a different kind of consideration. Bayer has now paid or committed to pay in the region of US$18 billion to resolve claims from people who say Roundup gave them cancer. The legal basis of most of those claims was not that glyphosate definitively causes non-Hodgkin lymphoma. It was that Monsanto knew or should have known there was a risk, and failed to warn users adequately on the product label.
Juries repeatedly found that standard met. Which means that even without a formal admission of liability, the litigation has established one significant finding: that a foreseeable risk existed, that users were not adequately warned, and that a company worth tens of billions of dollars judged it preferable to pay those sums rather than defend its product’s safety record in front of juries.
That is not a legal admission. But it is something.
In a June 25, 2026, development, the US Supreme Court ruled in Monsanto’s favour in the case of Monsanto v. Durnell, finding that federal pesticide law — which requires the EPA to approve product labels — preempts state-level failure-to-warn claims. The practical effect is that future plaintiffs will find it significantly harder to sue over labelling in state courts. It is a legal ruling about federal preemption, not a scientific ruling about whether glyphosate is safe. The underlying health questions are unchanged.
“Used According to the Label”: Why That Phrase Doesn’t Protect Most People
Central to the EPA’s reassurance — and to regulatory positions in New Zealand and elsewhere — is the phrase “when used according to the label.” No risks of concern, the assessment says, when glyphosate is used as directed.
Look more closely, and the phrase starts to unravel.
“Used according to the label” assumes that every user reads every instruction. That protective equipment is worn correctly, every time. That application rates are precisely followed. That there is no spray drift beyond the target area. That bystanders, children, neighbouring properties, and waterways are not affected. That exposure is a single, controlled, intentional event — not cumulative across years of farming, contracting, gardening, or simply living near treated land.
None of those assumptions hold consistently in the real world.
Spray drift is not a theoretical concern. It is a documented, measurable phenomenon that carries glyphosate beyond the field boundary into residential areas, schools, playgrounds, and waterways. Burst spray packs, equipment failures, and weather changes during application are not covered by label compliance. Farmworkers who mix and apply glyphosate-based products over seasons and years are not experiencing the exposure scenario the label assessment models.
And there is a further problem the label assessment does not address: commercial glyphosate formulations are not pure glyphosate. Products like Roundup contain adjuvants and surfactants — compounds that help the herbicide penetrate plant tissue. Research suggests that some of these formulations are significantly more toxic than glyphosate tested in isolation. The regulatory safety assessment is largely built on data for the active ingredient alone.
Two Standards, One Population: How Regulatory Language Shifts the Burden onto Consumers
The regulatory picture that emerges from all of this is one of genuine complexity — but also of consistent patterns.
IARC, using only publicly available independent research, found glyphosate probably carcinogenic. The EPA, using a broader dataset that included unpublished industry studies, found it not likely carcinogenic — a conclusion a federal appeals court subsequently found to be procedurally and scientifically flawed. A decade of litigation has resulted in nearly US$18 billion in settlements from a company that continues to deny its product causes cancer. And New Zealand, while setting stricter domestic residue limits and banning pre-harvest desiccation, has left a trade agreement in place that allows imported food to exceed those limits without consequence.
At each point where a stronger protective standard might have been applied, a different path was taken. And at each point, the burden of that decision falls not on the companies that manufacture glyphosate-based products, not on the regulators who assess them, and not on the legislators who set the framework.
It falls on the people who eat the food. Who live near the fields. Who mix and apply the product. Who breathe the drift.
The language of regulation — “probably,” “not likely,” “used according to the label,” “legally sold in Australia” — is precise in its own terms. But precision in language is not the same as protection in practice. And the gap between those two things is where the public health question actually lives.
No More Glyphosate NZ (NMGNZ) advocates for the restriction of glyphosate to genuinely appropriate uses, and for food standards that reflect New Zealand’s own residue limits — not those of our trading partners.
Key sources referenced in this article:
- IARC Monograph on Glyphosate, 2015 — Group 2A classification and basis
- US EPA Glyphosate page — “not likely to be carcinogenic” assessment
- Ninth Circuit Court of Appeals, 2022 — ruling that EPA’s assessment was procedurally and scientifically flawed
- Roundup litigation overview — Wisner Baum — settlement figures and case history
- US Supreme Court — Monsanto v. Durnell, June 2026 — federal preemption ruling on labelling claims
- NMGNZ — Glyphosate in Breakfast Foods: Weet-Bix and sorghum test results — independent testing and TTMRA implications
- NMGNZ — What If Our ‘Natural’ Food Isn’t? — Trans-Tasman Mutual Recognition Agreement analysis
- NMGNZ — “No Health Risk”? What MPI Didn’t Explain About Glyphosate Limits — analysis of MPI’s MRL review outcome
- MPI media release — Glyphosate residue limits to stay at 0.1 mg/kg for wheat, barley and oats — official November 2025 announcement
- NZ MPI — Trans-Tasman Mutual Recognition Arrangement — official framework
- MPI — Summary of Submissions on Proposed Glyphosate MRL Amendments, 2025 — MRL review and 3,100 submissions
Image Source & Attribution
We’re grateful to the talented photographers and designers whose work enhances our content. The feature image on this page is by Pixavril.
