Originally published on www.sayerji.substack.com
A novel class of genetically engineered food has quietly colonized the American diet since 2017. It doesn't make a poison. It switches off genes. And there is no label telling you it's there.
Dinner, Disrupted
Picture a family eating dinner in the American Midwest. The kids are having corn chips and juice. Dad's eating a pork chop. Mom reaches for the salsa -- corn-based, like nearly everything on the table. It's an ordinary evening in the most ordinary country in the world.
Now consider what might be happening at the molecular level inside that corn.
Embedded in the genome of the corn plants that grew those chips -- the same corn that makes the high-fructose syrup in that juice, the same corn silage that fed that pig -- is a genetic sequence encoding a molecule called double-stranded RNA, or dsRNA. Think of dsRNA as a search-and-destroy code for genes: a precisely designed molecular message that, when read by a cell's machinery, identifies a target gene and shuts it down. The target, in this case, is a gene in the western corn rootworm -- a devastating agricultural pest. The technology is called RNA interference, or RNAi. The corn is called SmartStax PRO, and it was approved by the EPA in June 2017.
Since that near-silent approval, RNAi crops have expanded to somewhere between 20 and 30 million acres of American farmland. A sprayable version -- a gene-silencing aerosol applied directly to potato fields -- has been commercially available since 2024. A second sprayable product, this one fed directly to honeybee colonies to kill varroa mites, received federal registration in September 2025. RNA-based fungicides are now before regulators in the US, the EU, and Brazil. An RNA herbicide that silences genes in weeds rather than insects is moving through development.
The family at the dinner table knows none of this. There is no label. There is no mandatory disclosure. There is, to all appearances, just dinner.
This is the story of how we got here, what the science actually says, and why the window to demand a different answer is closing faster than anyone in Washington will admit.
What RNAi Is -- and Why It's Different From Every GMO Before It
To understand why this moment matters, you need to understand what makes RNAi agriculture genuinely unprecedented.
Every GMO that came before it -- herbicide-tolerant soybeans, Bt corn, Golden Rice -- worked by adding or deleting a protein. The engineered gene produced a new protein, or it knocked out a protein the plant already made. Proteins are the workhorses of biology, and their effects, while sometimes complex, can in principle be tested with standard toxicological tools: feed it to rats, measure the outcomes.
RNAi works at an entirely different level of biology. It doesn't produce a protein at all. It produces RNA -- specifically, double-stranded RNA -- which acts not as a building block but as an instruction. RNA is the software running on the DNA hardware of a cell. And double-stranded RNA, in particular, is something mammalian biology has evolved to treat with extreme caution.
Here's why. Long before the first drug or pesticide was ever invented, viruses were invading cells and replicating using double-stranded RNA as part of their life cycle. Mammalian immune systems evolved a hair-trigger response to any long, perfectly duplexed RNA in the cellular environment, treating it as a viral alarm signal -- what immunologists call a PAMP, or Pathogen-Associated Molecular Pattern. When the immune system detects a PAMP, it doesn't pause to sequence it or identify its source. It responds: shutting down protein synthesis, degrading cellular RNA, triggering interferon, and, in severe or sustained cases, initiating cell death.
The DvSnf7 dsRNA in SmartStax PRO corn -- the active genetic payload in the world's first commercial RNAi crop -- is a 240 base-pair hairpin. The threshold for triggering the mammalian interferon response is 30 base pairs. The threshold for maximal activation of PKR, the immune kinase that halts all protein translation in an affected cell, is 85 base pairs. SmartStax PRO's payload clears both thresholds by a wide margin.
This is not a theoretical concern invented by critics. It is established molecular biology, documented in peer-reviewed literature, and identified in the canonical risk analysis of RNAi crops by independent scientists Jonathan Latham and Allison Wilson of the Bioscience Resource Project. Their framework -- which I first reported on in my 2015 GreenMedInfo article on the EPA's silent approval -- identified three distinct mechanisms by which RNAi crops could harm non-target organisms, including humans. Nearly a decade later, none of the three mechanisms has been adequately addressed by regulators.
And meanwhile, the food supply has been transformed.
The Gene That Never Asked Permission
The story of how RNAi corn entered American agriculture without public knowledge begins with a regulatory decision in June 2017 that Bill Freese of the Center for Food Safety, reporting to The Atlantic, called a masterpiece of bureaucratic invisibility.
The EPA offered just 15 days of public comment on SmartStax PRO -- compared to the standard 60 days for significant regulatory actions. The agency did not post the approval to the Federal Register, the standard mechanism by which the public and Congress are notified of major regulatory decisions. The technology landed in American cornfields with the quiet authority of a memo circulated among insiders -- which is essentially what it was.
SmartStax PRO was not a simple product. As I documented in that 2015 article, the approved stack contained six separate modes of action stacked into a single corn variety: the DvSnf7 dsRNA RNAi payload, three separate Bt proteins for rootworm and above-ground pest control, glyphosate tolerance, and glufosinate tolerance. This complexity was not incidental. Embedding the RNAi trait within an already-complex toxicological cocktail made it impossible to isolate RNAi-specific effects in any subsequent field monitoring. If something went wrong in the populations eating this corn, identifying dsRNA as the cause would require a level of molecular epidemiology that the United States has never performed for any agricultural biotechnology.
The EPA's own Scientific Advisory Panel had flagged "ongoing areas of uncertainty" in its 2016 meeting minutes -- concerns that dated back to a 2014 SAP meeting and had not been resolved. The approval went forward anyway.
No new RNAi-specific SAP meeting appears on the EPA's advisory panel list for 2025 or 2026. The scientific review has not kept pace with the commercial expansion.
Since 2017, Bayer has steadily expanded SmartStax PRO acreage. By the 2024 growing season, Bayer's seed lineup featured 25 new DeKalb hybrids carrying SmartStax PRO or the related VT4PRO RNAi technology -- a second Bayer trait launched commercially in Eastern Canada in 2025. Corteva launched its own competing RNAi corn product, Vorceed Enlist, in 2023, targeting the same DvSnf7 gene sequence in corn rootworm. Both companies' RNAi acres are additive. The total US footprint of RNAi corn in 2025 is conservatively estimated at 20 to 30 million acres -- roughly 20 to 30 percent of the total US corn crop.
That corn is the genomic backbone of approximately 75 percent of processed foods sold in American supermarkets.
Food as Information: What Science Has Quietly Established
To grasp what is truly at stake, you have to understand a revolution in biology that has been unfolding in the scientific literature for the past two decades -- a revolution that most people, including most physicians, have not yet absorbed.
The human genome, as I explored in "Genetic Dark Matter and the Return of the Goddess", is not primarily a protein-coding machine. Only about 1.5 percent of human DNA encodes proteins. The remaining 98.5 percent -- dismissed for decades as "junk DNA" -- produces a vast landscape of regulatory RNA molecules, including microRNAs, that govern the expression of approximately one-third of the entire protein-coding genome. The Human Genome Project's revelation that humans possess only around 23,000 genes -- fewer than a grain of rice -- was a paradigm-shattering moment: complexity lives not in gene count but in the regulatory RNA environment.
This regulatory landscape is not hermetically sealed inside the human body. Plants also produce microRNAs. And as a landmark 2012 study in Cell Research by Chen-Yu Zhang and colleagues found, plant microRNAs -- specifically, a rice microRNA called miR168a -- were detectable in the serum of Chinese subjects who ate rice-rich diets. More than that: miR168a appeared to suppress the expression of a liver gene called LDLRAP1, which governs LDL cholesterol clearance. Food wasn't just fueling the body. Food was talking to the genome.
As I documented in "The Dark and Light Side of Food As Information", this was not an isolated finding. Plant miR2910 -- conserved across fruits and vegetables humans have eaten for hundreds of thousands of years -- appears in high relative abundance across 410 human plasma small RNA sequencing datasets. It shares structural features with human microRNAs and is predicted to modulate the JAK-STAT signaling pathway through its target gene SPRY4, a pathway governing immunity, cell differentiation, proliferation, and the suppression of cancer. The emerging picture is one of cross-kingdom nutritional signaling: plant-derived microRNAs may constitute an evolutionarily ancient information channel through which the human genome receives calibrating molecular signals from its food environment.
A 2025 review in Frontiers in Nutrition confirmed this evidence base, describing how plant miRNAs can "survive digestion, enter the mammalian circulation, and modulate host gene expression to influence glycolipid homeostasis," and identified SIDT1 -- a transmembrane protein -- as a candidate active uptake mechanism. The science has moved beyond passive diffusion.
Is this universally accepted? No. A rigorous September 2025 systematic review in Advances in Nutrition by Tambaro and colleagues -- the most careful assessment to date -- found methodological heterogeneity that explains much of the positive signal, and concluded that "no direct molecular evidence currently demonstrates that plant xenomiRs bind to mammalian silencing machinery." A large-scale analysis of 824 human datasets found xenomiRs -- foreign plant microRNAs in human blood -- were "likely artifacts" from technical contamination. A 2024 review in ExRNA, however, found "increasing evidence suggests these dietary miRNAs are not only capable of being absorbed by consumers such as humans, but also appear to be extensively involved in various physiological activities."
Both sides agree on the most important point: the question is genuinely open.
And that is precisely why what is being inserted into this signaling system -- without consent, without labels, without long-term study -- matters so much.
The Dark Side: Novel Instructions in an Ancient System
The light side of the food-as-information story is this: the microRNAs in the plants you eat may be part of an evolutionary conversation between species, one that has been calibrating human gene expression for as long as we have been eating plants. The diet of your ancestors was, in this sense, a molecular curriculum.
The dark side is this: we are now inserting novel molecular instructions -- dsRNA sequences derived from insect genes, never part of any plant's evolutionary repertoire, never part of any human dietary history -- into those same food plants. And we are doing it without telling anyone.
The DvSnf7 dsRNA in SmartStax PRO corn is a 240 base-pair hairpin derived from the Snf7 gene of the western corn rootworm (Diabrotica virgifera virgifera). It is expressed constitutively -- throughout the growing season, in roots, leaves, and pollen -- in corn plants that now occupy tens of millions of American acres. It is not a sequence that appears anywhere in the evolutionary dietary history of Homo sapiens. It is a novel molecular instruction, inserted into the most prevalent food crop in the American diet, without the knowledge or consent of the people eating it.
The Calantha dsRNA (ledprona) -- targeting the PSMB5 gene in the Colorado potato beetle -- is a 490 base-pair sequence sprayed onto potato plants in the field. Potatoes are consumed with minimal processing: baked, boiled, fried. Unlike corn, which undergoes extensive industrial transformation into syrups and starches and chips, the potato on your plate is close to the field. No published study has examined whether ledprona residues survive in commercially cooked potatoes. GreenLight Biosciences secured Calantha's final EPA registration in January 2024 and the product was commercially available to potato growers by spring of that year.
Monsanto's own 2008 research -- cited in my 2024 GreenMedInfo investigation -- found that endogenous corn small RNAs already match approximately 450 to 2,300 unique RNA transcripts in rat, mouse, and human genomes. Industry interpreted this as evidence that mammals safely consume plant RNA. Critics, including Latham and Wilson, read it differently: it demonstrates that plant small RNAs are already complementary to mammalian genes, meaning that deliberately engineering plants to over-express specific dsRNA sequences of defined length and complementarity doesn't introduce a new risk category from scratch. It adds an intentional layer of cross-species gene-targeting onto an already present background of natural cross-kingdom RNA signaling -- with no safety testing to characterize the interaction.
This is the question science has not answered. This is the experiment now running on 330 million Americans.
What Bayer Doesn't Print on the Box
In the regulatory universe that approved these products, the standard of safety assessment for SmartStax PRO's dsRNA payload rested primarily on a single study: a 2016 repeat-dose oral toxicity study in mice, published in Regulatory Toxicology and Pharmacology and funded by Monsanto, now Bayer, that found no treatment-related effects at doses up to 100 milligrams per kilogram -- millions to billions of times higher than any anticipated human dietary exposure.
What that study did not do: it used direct stomach delivery (gavage), which bypasses mastication and normal gut processing. It tested the intact 968-nucleotide RNA molecule, not the shorter siRNA fragments that might persist in processed foods. And critically, it did not assess immunotoxicity via the PAMP pathway -- the innate immune alarm system activated by long dsRNA regardless of its sequence. You can give an animal a dose of dsRNA vastly exceeding any real-world exposure, see no obvious toxicity in the standard endpoints you measure, and completely miss the molecular effects that the PAMP mechanism would produce in a gut epithelial cell encountering the molecule in vivo.
The bioinformatics framework for evaluating human off-target risk from sprayable dsRNA pesticides -- the basic analytical tool for asking "could this dsRNA silence genes it wasn't designed to target in humans?" -- was not published until March 2025. Researchers at Johns Hopkins proposed this first standardized methodology in a paper that explicitly acknowledged it was being developed for evaluating Calantha/ledprona. SmartStax PRO had been on the market for eight years. Calantha had been commercially sold for a full growing season. The analytical framework for asking the most basic safety question arrived after the products.
A parallel 2025 paper in Integrated Environmental Assessment and Management confirmed that as of 2025, the EPA had reviewed bioinformatic analyses for only four dsRNA products -- and that the evaluation standards recommending "21-nucleotide stretches of perfect identity and 80% overall identity" between dsRNA and non-target transcripts had never been codified into regulation. These are guidelines. They are not law. Companies can and do deviate from them.
And in Europe, the structural problem is even more pointed. The European Food Safety Authority published a review of RNAi risk assessment methodology in December 2024 revealing that its new bioinformatics tool for off-target analysis -- the core analytical safety instrument -- "will not be performed in the EFSA environment," is "to be run by the applicants," and that "the outcome... is owned by the applicant." The corporation seeking approval runs the safety analysis and owns the data. EFSA sees what the company chooses to show it.
This is not a regulatory framework. It is a regulatory theater performed by the regulated parties, for an audience that will never see the script.
1,400 Bees and a Silence That Still Hasn't Been Answered
The most striking single experiment in the RNAi crop safety literature involves honeybees -- and its implications have never been refuted.
In 2015, researchers published a study in the Journal of Immunology Research examining what happened when honeybee larvae were exposed to dsRNA -- not DvSnf7, but a control sequence targeting green fluorescent protein (GFP), a molecule entirely foreign to bees. The logic of a GFP control is straightforward: if sequence-specific RNAi is the only mechanism at work, dsRNA targeting GFP should have no effect on bees whatsoever, because bees have no GFP gene to silence.
What the researchers found was not what the industry model predicts.
Feeding honeybee larvae the GFP-targeting dsRNA produced approximately 1,400 differentially regulated genes -- meaning 1,400 bee genes that changed their expression pattern in response to the dsRNA. Of those 1,400 genes, only nine shared any sequence similarity with the GFP dsRNA. The other 1,391 gene expression changes were entirely sequence-independent: triggered not by the specific molecular message the dsRNA carried, but by the fact of its being dsRNA at all.
This is the PAMP mechanism in action. The bees' innate immune systems read the dsRNA as a viral alarm signal and responded accordingly -- activating immune and stress pathways wholesale, without regard to sequence. The result was a cascade of biological changes that conventional toxicology endpoints would not detect, but that could easily impair colony health, navigation, foraging behavior, or immune resilience under field conditions.
This finding has never been superseded by industry-sponsored studies, which have focused narrowly on lethal endpoint assays -- asking "does the bee die?" rather than "what does dsRNA exposure do to bee biology?" As Latham and Wilson noted, these are profoundly different questions.
Now consider the newest product in the RNAi agricultural portfolio: Norroa (vadescana), manufactured by GreenLight Biosciences and granted EPA registration in September 2025. Norroa is designed to protect honeybee colonies from varroa mites -- a genuine threat to global pollinator health. The product is fed to bees in hive strips. The forager bees that metabolize it travel to flowers, collect pollen and nectar, return to the hive, and produce honey.
Has any published study examined whether vadescana or its siRNA fragments persist in commercial honey? No. Honey is consumed without processing by millions of people, including infants and children. The EPA review found no RNA sequence matches between vadescana and honeybee RNA -- but the 1,400-gene bee study shows that the absence of sequence matches does not rule out sequence-independent immune effects. This is the most direct and most underexamined human dietary exposure pathway for any registered dsRNA product currently on the market.
If you ate honey this week, you are in the experiment.
The Infant at the Table
The family at the dinner table has a baby in a high chair. She is six months old and drinking formula.
That formula almost certainly contains corn starch or corn syrup solids derived from American field corn -- the same corn supply that now carries RNAi traits across 20 to 30 million acres. The corn starch in infant formula is wet-milled from field corn. It may or may not carry intact dsRNA. It has not been systematically tested.
Latham and Wilson specifically identified two populations as facing elevated risk from RNAi crop exposure: newborn infants and adults with inflammatory bowel disease, Crohn's disease, or other conditions causing intestinal permeability.
The reasoning is physiological. In a healthy adult, the gastrointestinal tract presents multiple barriers to macromolecular absorption: stomach acid, pancreatic RNA-degrading enzymes (ribonucleases), and the tight-junction architecture of intestinal epithelial cells. These barriers may degrade most or all ingested dsRNA before it reaches the bloodstream. The industry's safety case for SmartStax PRO corn rests heavily on this assumption.
But newborn infants do not have intact gut barriers. The neonatal intestine is physiologically permeable -- intentionally so, to allow maternal antibodies from breast milk to pass into circulation and confer early immune protection. The enzymatic complement of the infant gut is immature. The innate immune system is still developing. The very features that make the infant gut biologically appropriate for absorbing protective molecules from breast milk also make it more permeable to molecules that a mature adult gut would block.
No regulatory body has required a safety study for infants consuming formula derived from RNAi corn starch or corn syrup. The Latham-Wilson paper flagged this gap in 2015. As of 2026, no such study exists in the public literature. An FOIA request to the FDA or EPA asking for any infant exposure assessment data related to DvSnf7 dsRNA would almost certainly return null results. The government has never asked the question.
This is not a regulatory gap. It is a regulatory choice.
The Beekeeper's Hive and the Monarch's Field
A beekeeper in Iowa checks her hives in July. The surrounding fields are planted in DeKalb hybrids carrying SmartStax PRO. Her bees forage on surrounding wildflowers and on crop corn pollen throughout the growing season.
The corn pollen of transgenic RNAi crops contains the DvSnf7 dsRNA payload. Pollen-borne dsRNA was documented in a related Bayer trait (MON84711), establishing that pollinators are exposed through foraging in RNAi corn landscapes. What does that exposure do to bees? The 1,400-gene study suggests the PAMP mechanism is a real biological risk. But the EPA's testing framework focused on standard lethal toxicity endpoints. The beekeeper has no way of knowing whether her colony's immune resilience, navigation, or foraging efficiency has been affected by dsRNA exposure from surrounding fields. The data simply doesn't exist.
The monarch butterfly faces a parallel silence. Monarch populations have declined precipitously over the past three decades, correlating with the expansion of GMO corn and soy acreage across the Corn Belt. The primary documented mechanism has been glyphosate-driven milkweed elimination. But the parallel expansion of RNAi corn through the same agricultural landscape -- and the documented sequence-independent PAMP effects of dsRNA on insect biology -- has never been examined as a potential contributing factor. Laboratory studies, including a 2025 study in Insects examining the varroa mite dsRNA vadescana, found that lepidopterans (the order that includes monarchs) are generally refractory to oral RNAi through the standard gene-silencing mechanism. The authors note this likely reflects a biological characteristic of the order rather than proof that dsRNA is inert in these organisms.
Sublethal immunotoxicity in monarchs from chronic low-dose dsRNA pollen exposure in a landscape saturated with RNAi corn has never been studied.
We don't know because we haven't asked.
A 2022 review in Science of the Total Environment summarized the state of knowledge on RNAi off-target effects in insects: "there is a clear lack of consensus on the minimum necessary sequence similarity requirements for effective RNAi in insects," and cross-species off-target prediction using current bioinformatic tools remains "not possible." The technology is commercially deployed across tens of millions of acres. The scientific tools to evaluate its ecological safety don't yet exist.
The Long Food Chain: From Corn to Cow to Your Plate
RNAi corn doesn't just become chips and soda. Most of it travels a longer and less visible path into the American diet.
The USDA Economic Research Service estimates that approximately 45 percent of the US corn crop goes to fuel ethanol production -- but that process generates a massive co-product: distillers' dried grains with solubles, or DDGS, which is fed in enormous quantities to dairy cattle, beef cattle, pigs, and poultry. Another 40 percent goes to direct livestock feed. The 15 percent allocated to food, seed, and industrial uses becomes the high-fructose corn syrup in your soda, the corn starch in your baked goods, the corn oil in your salad dressing, the corn flour in your tortilla.
Of the approximately 86 million GMO corn acres in the United States, a conservative 20 to 30 percent now carry RNAi traits. Run that through the livestock supply chain and a realistic estimate puts 8 to 12 percent of total US corn production passing through RNAi-trait plants into animal feed consumed by the dairy and beef and pork industries before arriving on American plates as meat, dairy, and eggs.
The DDGS pathway deserves particular attention. Ethanol fermentation of RNAi corn involves yeast processing at temperatures that may partially -- but not completely -- degrade dsRNA. The DDGS that emerges from distillation is higher in protein concentration than the input grain, because the fermentation process removes starch while concentrating everything else. Whether that "everything else" includes elevated RNA concentrations -- from both the plant-derived dsRNA payload and from microbial RNA generated during fermentation -- has never been systematically investigated.
The Non-GMO Project estimates that corn derivatives appear in approximately 3,000 items in the average American supermarket. A person eating a typical American processed-food diet has, by any reasonable calculation, daily contact with ingredients derived from RNAi corn. Whether that contact involves intact or processed dsRNA fragments, whether those fragments survive gastric conditions to reach intestinal cells, and what they might do once they get there -- these are the critical unanswered questions at the center of this story.
The food industry cannot answer them because it has never been required to ask.
"Non-GMO" -- and Why That Label Is Now False
GreenLight Biosciences markets Calantha -- its sprayable dsRNA insecticide -- as "non-GMO."
The argument is technically coherent on its narrowest reading: Calantha is a pesticide applied to potato plants from the outside. The potato plant's own genome is not modified. Therefore, by the conventional definition of "GMO" as referring to a plant with modified genetic material, a potato treated with Calantha is not a GMO potato.
But the Friends of the Earth 2020 report -- authored by Dana Perls, who has been tracking this issue as closely as anyone in the environmental policy world -- documented what the "non-GMO" framing omits: exposed organisms -- insects, soil microbes, potentially surrounding plants -- may have their gene expression altered by the dsRNA. GreenLight Biosciences simultaneously holds patents on those dsRNA sequences. A patent on a gene-silencing molecule used in an open field creates the same IP landscape that Monsanto successfully weaponized against farmers whose crops were contaminated by GMO pollen: if the dsRNA drifts to a neighboring field and is absorbed by a plant, theoretically altering its gene expression, the question of who owns the resulting organism is not merely theoretical. It is a legal time bomb.
The FOE report stated explicitly that "biotech companies filing patents that claim property rights over exposed organisms and their offspring" constitutes a primary socioeconomic harm of RNAi pesticide deployment. No legal test case has been litigated. But the IP framework enabling this scenario is already in place.
Perls put it directly: "It is not currently possible to assure the safe use of RNAi products designed to induce genetic modifications in organisms in the open environment."
And yet the product is on shelves, marketed as "non-GMO," with no label identifying the potato in your grocery store as having been treated with a gene-silencing spray.
The SECURE Rule Falls, the Window Closes
In December 2024, a federal court in the Northern District of California vacated the USDA's 2020 SECURE Rule -- the modernized framework that had updated biotech crop oversight to focus on organism properties rather than the process used to create them. The ruling reverted USDA oversight to pre-2020 standards.
This might sound like a tightening of regulation. It wasn't. The SECURE Rule's "Am I Regulated?" pathway had been used by developers to determine whether new crops required full regulatory review. Its vacating created not more oversight but more uncertainty -- the kind of uncertainty that commercial actors know how to navigate and that public interest advocates do not. USDA restarted its permitting processes in early 2025, but acknowledged that any crops reviewed under the SECURE Rule between 2020 and December 2024 remain valid.
One month before the SECURE Rule was vacated, in October 2025, the Ninth Circuit Court of Appeals found that the USDA had improperly excluded refined foods -- high-fructose corn syrup, corn oil, corn starch -- from the mandatory "bioengineered" disclosure requirement on the grounds that modified DNA is rendered undetectable in the final product. The court held this exclusion was legally improper. But it left the underlying regulations in place pending USDA revision. Until that revision happens, the corn syrup in your soda, the corn starch in your baby's formula, and the corn oil in your salad dressing carry no disclosure -- even if they are derived from RNAi corn.
Consumer advocates at the National Sustainable Agriculture Coalition and Beyond Pesticides have called for expanding the National Bioengineered Food Disclosure Standard to include CRISPR and RNAi techniques. As of writing, that expansion has not occurred. Sprayable RNAi pesticides -- Calantha, Norroa -- are not covered by the NBFDS at all, because they are classified as pesticide residues, not bioengineered food ingredients. A potato treated with a gene-silencing spray has no required label indicating this fact.
The "bioengineered" label on the products that do carry it uses legally mandated language that courts have found misleading because most consumers do not recognize "bioengineered" as synonymous with "genetically modified." A consumer scanning a corn chip bag for the word "GMO" will not find it. They will find, in small type somewhere near the ingredients panel, a phrase their brain has no framework to decode.
This is not an accident. It is a design.
The Australian Exception
There is one moment in the regulatory history of RNAi crops where the system worked as it was supposed to -- and it happened not in the United States, but in Australia.
CSIRO, Australia's federal scientific research agency, developed an RNAi wheat designed to produce a lower-glycemic flour by silencing the waxy starch synthase gene. It was a compelling application -- a wheat engineered not to produce a new toxin but to alter its starch composition.
Then Jack Heinemann of the University of Canterbury and his colleagues performed a bioinformatics analysis, published in Environmental International in 2013, examining whether the siRNA fragments generated by the wheat's RNAi pathway shared sequence similarity with any human gene. The analysis found that the silencing sequences targeting the plant waxy starch synthase gene shared similarity with the human glycogen branching enzyme gene -- GBE. A deficiency in GBE causes glycogen storage disease type IV, or Andersen's disease, a condition that can be fatal in children. As documentation of Heinemann's analysis noted, the bioinformatics could not rule out unintended cross-reactivity.
The CSIRO wheat was not commercialized. In this one case, the precautionary principle prevailed.
Compare that to the American regulatory track record: 15 days of public comment, no Federal Register notice, no PAMP assessment, no infant formula study, no long-term feeding studies in mammals, no independent non-target organism testing beyond standard lethal endpoints. SmartStax PRO entered 15 million acres -- now expanding toward 20 to 30 million -- without any of the analysis that kept CSIRO's wheat off Australian supermarket shelves.
Australia asked the question. The United States never did.
The Clinical Warning Everyone Ignored
If you want to understand the asymmetry between how the drug industry treats RNAi risk and how the agriculture industry does, consider what happened to Alnylam Pharmaceuticals in 2016.
Alnylam was developing revusiran, an RNAi therapeutic for a rare and fatal heart condition. The molecule was specifically engineered for a human therapeutic target, designed by some of the world's leading RNA scientists, reviewed under the FDA's rigorous drug approval process, and tested in Phase 3 clinical trials in a controlled population.
In October 2016, Alnylam discontinued the trial after an observed mortality imbalance: 18 deaths -- a 12.9 percent mortality rate -- in the revusiran arm versus 2 deaths, a 3.0 percent rate, on placebo during the on-treatment period. Alnylam's stock fell nearly 50 percent. A 2020 post-hoc analysis failed to identify a definitive pharmacological cause.
This remains the most direct clinical evidence that RNAi-based molecular interventions can cause unexpected mortality at scale -- even when the dsRNA is precisely engineered for a therapeutic target, with extensive regulatory review, in a highly controlled setting.
In agricultural RNAi deployment: the exposed population is not a carefully selected clinical trial cohort. It is the entire American population, including the most vulnerable individuals -- infants, the elderly, the immunocompromised, those with inflammatory bowel disease or intestinal permeability. The safety review is not the FDA's drug approval process. It is a 15-day public comment period and a bioinformatics analysis run by the applicant.
The lesson of revusiran for agricultural RNAi has never been formally addressed by any regulatory body.
The Epigenetic Long Game
Here is the dimension of this story that makes even the most cautious scientists uncomfortable when they think about it carefully.
A 2024 review in Trends in Plant Science documented that siRNAs -- the short RNA fragments produced when dsRNA is processed by the cellular machinery -- can induce epigenetic changes, including a process called RNA-directed DNA methylation (RdDM), that persist for up to nine generations in plant hybrid lines. The dsRNA doesn't just silence a gene temporarily. Under the right conditions, it can alter the chemical marks on DNA itself, creating changes that are heritable across generations without any alteration to the underlying DNA sequence.
In March 2026, separate studies in Critical Reviews in Toxicology and PNAS documented pesticide-induced epigenetic changes in mammals persisting across 20 generations -- transmitted through both maternal and paternal lineages. The regulatory frameworks governing pesticide safety focus on direct toxicity: does the animal die, does the organ fail, does the tumor grow? They were not designed to detect epigenetic modifications propagating quietly across generations.
Whether repeated dietary exposure to RNAi-derived food could induce heritable epigenetic modifications in humans has never been studied. No one has required that it be studied. The 20-generation rodent timescale corresponds, in human terms, to roughly 400 years of potential effect propagation. By 2026, we have had nine years of commercial RNAi corn in the food supply. The epidemiological signal from a multigenerational epigenetic disruption -- if one exists -- will not appear in any dataset that exists today.
This is precisely what makes the precautionary principle not a luxury but a necessity: we are operating in a regime where the consequences of error, if error there be, will be invisible to science for decades after the point of irreversibility.
The Window Is Closing
In October 2025, the global RNAi pesticides market was valued at USD 1.2 billion, with projections reaching USD 4.6 billion by 2034 -- a compound annual growth rate of 14.2 percent. Fifteen major corporations are in active development across insecticide, fungicide, and herbicide applications. In February 2025, the European Investment Bank committed €35 million to GreenLight Biosciences Españafor a pipeline of ten EU products. In May 2026, GreenLight secured the first RNA-based insecticide emergency use authorization in the European Union -- a regulatory beachhead from which broader European market entry becomes inevitable.
In April 2025, GreenLight announced a breakthrough in an RNA herbicide platform targeting horseweed, with a pipeline of 180 candidate dsRNA sequences selected using AI-enabled design tools. A dsRNA sequence targeting a plant's own gene expression -- rather than an insect's -- shares evolutionary ancestry with the plant kingdom that humans eat, and carries a fundamentally different and higher-order cross-kingdom risk profile than insect-targeting sequences. No public risk assessment for plant-targeting RNA herbicides exists.
In October 2025, GreenLight submitted regulatory dossiers simultaneously in the US, EU, and Brazil for a dsRNA fungicide targeting grape powdery mildew -- which would bring RNA gene silencing directly into wine grapes and the wine and juice supply for the first time.
Bayer launched 25 new DeKalb hybrids with SmartStax PRO or VT4PRO RNAi technology in 2024. The VT4PRO commercial launch in Eastern Canada in 2025 signals that RNAi is no longer a niche Bayer product -- it is the core of the company's corn trait strategy going forward.
The window for meaningful democratic intervention in this technology -- mandatory labeling, independent long-term safety studies, PAMP-specific risk assessment for vulnerable populations, a coherent regulatory framework for nanoformulated dsRNA products designed to resist the rapid environmental degradation that industry currently cites as its primary safety argument -- is not a theoretical future concern. AgroSpheres, Pebble Labs, and RNAissance Ag are actively developing nanoparticle delivery systems for dsRNA using chitosan, lipid nanoparticles, star polycations, and clay nanosheets, explicitly designed to increase environmental persistence and cellular uptake. The same rapid degradation data that industry cites to dismiss persistence concerns does not apply to encapsulated formulations. No public regulatory guidance addresses this distinction.
By the mid-2030s, at the current trajectory, RNA-silencing pesticides will be as commercially embedded in the global food system as Bt crops and glyphosate are today. Regulatory review at that point becomes not prevention but remediation -- and remediation of an invisible, unlabeled, RNA-level intervention in the food supply has no historical precedent.
The window is 2026 to 2030. After that, the experiment becomes permanent.
The 500-Million-Year Conversation
I want to close not with a policy prescription but with a frame -- because the frame matters enormously for what comes next.
The cross-kingdom RNA signaling that makes this story scientifically significant is not a laboratory curiosity. As I explored in "Genetic Dark Matter and the Return of the Goddess", and as the accumulating xenomiR literature continues to suggest -- with all its genuine uncertainty and methodological disputes -- the molecular conversation between the plants we eat and the genomes we carry is old. It is older than agriculture. It is older than Homo sapiens. It is a conversation that has been running, through every meal eaten by every ancestor in every ecosystem since the first organisms began consuming other organisms for energy.
The human genome's ~98.5 percent non-coding fraction is not junk. It is the accumulated regulatory wisdom of 500 million years of that conversation: microRNAs calibrated to dietary plants, epigenetic marks responsive to seasonal food cycles, immune responses tuned to the molecular signatures of the organisms in our environment. The holobiont -- the human body understood not as a discrete individual but as an ecological community of cells, microbiome, and dietary molecular signals -- is the unit that evolution actually selected. It is not the gene. It is not the individual. It is the conversation.
Into this conversation, commercial interests have now inserted novel molecular instructions derived from insect genes, never part of any plant's evolutionary history, never part of any human dietary inheritance. They have done so without asking permission -- not from the public, not from science, not from the regulatory process in any meaningful sense. They have done so at a scale that now encompasses tens of millions of acres, thousands of product derivatives, and a daily dietary exposure for most Americans.
And they have labeled it nothing.
The precautionary principle -- the simple idea that when a new technology might cause irreversible harm, the burden of proof falls on those introducing it, not on the public forced to absorb it -- is not a conservative impulse or an anti-science reflex. It is a recognition that some decisions, once made at scale, cannot be unmade. Glyphosate is in the rain. Microplastics are in human blood. The consequences of those introductions are still being counted. Adding a layer of novel gene-silencing RNA to the same food system, with less transparency and less regulatory scrutiny than either of those technologies received, is not progress. It is recklessness dressed in the language of precision.
As Friends of the Earth's Dana Perls has argued, as Latham and Wilson have argued, as Jack Heinemann demonstrated with the CSIRO wheat, as Beyond Pesticides called for in 2019, as more than 90 NGOs and the International Federation of Beekeepers' Associations wrote to the IUCN in 2025 to demand: no release should occur unless and until it can be demonstrated that there are no direct or indirect risks to pollinators, biodiversity, or ecosystems. That standard has not been met. It has barely been attempted.
The demand is not complicated. Label it. Study it independently. Give us a choice. Give the infant in the high chair a chance to opt out of the experiment she never consented to join.
The 500-million-year conversation between species that built the human genome deserves at least that much respect.
Sayer Ji is the founder of GreenMedInfo.com and has been investigating the intersection of food, genomics, and public health policy for over a decade. This article is an updated investigation building on his prior reporting: "Hidden Food Threat: Experts Warn of Dangers of RNAi Crops" (June 2024); "The GMO Agenda Takes a Menacing Leap Forward with EPA's Silent Approval of Monsanto/Dow's RNAi Corn"(2015); "Genetic Dark Matter and the Return of the Goddess"; and "The Dark and Light Side of Food As Information". Research period: literature and regulatory developments through May 2026.
No hay comentarios:
Publicar un comentario