(FORTUNE Magazine) – For anyone in the business of growing corn, one of the biggest frustrations of the job is a brown inchworm-like creature that spends most of the summer and fall munching and tunneling through the corn, only to emerge as a moth that flies off to spawn a lot more inchworms. Like many adolescents, corn borers can be enormously destructive. Depending on when in the growing season they arrive, they can damage arteries that carry moisture to the corn, or even cause the entire ear to fall off before harvest. The borer costs American farmers and their $20 billion corn crop upwards of $1 billion a year, if you count diminished yields plus the price of pesticides and other measures needed to keep the borer at bay.

So in 1995, when scientists produced an early variety of genetically modified corn that poisoned the borer shortly after its first cornstalk casserole, farmers fairly jumped for joy. But last summer, right in the middle of the harvest, things got messy. Plant Genetics Systems, a company now owned by Aventis, a giant European pharmaceuticals firm, had developed another borer-killing gene that it called Starlink. However, the toxin that Starlink produced in the corn plant resembled a substance that triggers violent allergies in some people. When federal regulators threatened to ban Starlink corn until its safety in humans could be established, the developers thought they had a better idea. In effect, they promised to sell Starlink seed only to farmers using it for feed corn; in turn, the farmers would agree not to sell the seed to anyone who would put it in human food. Okay, said the feds. But be careful.

Well, guess what? Almost everybody involved screwed up. Even though Starlink was on the market for just three years--and made up just 0.5% of the 80 million acres of corn planted in the U.S. last year--it began showing up in all sorts of places it didn't belong, including tacos, corn chips, breweries, and muffin mix. The promises made by Starlink's inventors proved worthless, falling prey to managerial inattention, corporate mergers, blind faith, misplaced hope, woeful ignorance, political activism, and probably greedy farmers too, if you can imagine such a thing.

The episode hardly qualifies as a disaster, since no one seems to have gotten seriously ill from eating Starlink corn. Howard Buffett, son of Warren and a farmer near Decatur, Ill., even sees a bright side to it; he says Starlink has revealed the shortcomings of federal oversight and has pointed up the inability of the grain-handling industry to segregate subtly different products. Still, Starlink has caused no end of hassles for farmers, grain-elevator operators, railroads, and food processors. Neil Harl, an agricultural economist at Iowa State University, calls it "the biggest assault on American agriculture I have ever witnessed." Altogether, the fiasco could cost Aventis half a billion dollars.

The long-term consequences may be more severe. So far Americans have been much more accepting of genetically modified food than the rest of the world. If Starlink triggers hysteria among Americans, the world's biggest appetite for that promising technology will shrink, and the whole science will be retarded for years. If foreign food processors that buy U.S. agricultural commodities worry that American grain glows in the dark, they will turn even more to Brazil and other countries for their food, and U.S. farm prices, already depressed, will fall further.

One of the more surprising revelations of the Starlink mess isn't that genetically modified food has suddenly appeared in the food supply, but rather how much such food is already out there. Most of us have heard about such oddities as strawberries protected from frost damage by a gene transplanted from an arctic fish. But did you know that genetically modified soybeans now account for 60% of all soy grown in the U.S.? Called Roundup Ready, the plants were developed by Monsanto to tolerate Roundup, one of the company's weed-killers. Says Gary Niery, a farmer in central Illinois: "Before Roundup, we used to use a quart of herbicide per acre. Now it's just ounces." Similarly engineered soy plants, including LibertyLink from Aventis, are sold by other companies.

Close on the heels of Roundup Ready soy came another kind of genetically altered plant: one that produced its own pesticide. That's where the Starlink story begins. For nearly 30 years farmers have sprayed crops with solutions derived from a soil bacterium called Bacillus thuringiensis. This so-called Bt spray is harmless to humans but quite effective against a variety of pests, including corn borers. However, it doesn't kill all corn borers, especially those that often show up in a second wave of infestation in midsummer. In 1995 seed companies such as Pioneer Hi-Bred and DeKalb won approval to sell corn genetically altered to produce the pesticide found in soil bacteria; this seed killed nearly 99% of corn borers. About 18% of corn planted in the U.S. last year was of the Bt variety.

One bag of seed corn (enough to plant 2 1/2 acres) costs $90; Bt corn costs an additional $15 per bag. Corn-borer infestations vary widely from year to year, depending on wind and rain. If infestations are mild, it's cheaper to fight the borer with sprays. But in broad swaths of the Cornbelt where the borer is a chronic problem, the Bt varieties of seed are more economical. Roughly a quarter of the corn grown last year in Iowa, Kansas, Nebraska, and Minnesota was of the Bt type; the figure was 35% in South Dakota.

Pioneer and DeKalb's head start in the Bt- engineered crop business worried Aventis, a $20-billion-a-year French pharmaceuticals and agricultural sciences company formed last year by the merger of Rhone-Poulenc and Hoechst. Although most of Aventis' revenues come from drugs such as Allegra, a prescription antihistamine, the company's crop-sciences division had sales of $4 billion last year, making it one of the biggest ag-products operations in the world. At least some Aventis officials had big hopes for genetic engineering. "We were spending $450 million a year on R&D in the agricultural division," says an executive who, like all Aventis officials interviewed by FORTUNE, declined to be identified. "We had gone about as far as you could fighting weeds and pests with chemicals and needed to make a big shift to biotechnology." Even before their companies merged, executives at Rhone-Poulenc and Hoechst worried that rivals were grabbing market share in key agricultural technologies that would be difficult to win back later.

Buried in the welter of corporate subentities created by the Rhone-Poulenc/Hoechst combination was a small Belgian company called Plant Genetics, which Hoechst had acquired in 1996. Corporate life cannot have been easy for the managers and scientists at Plant Genetics, who had been working for a decade on a Bt variety of corn. Four years before the Aventis merger, Hoechst had formed a joint venture with Schering, the U.S. drug company. Plant Genetics was acquired by the joint venture, called Agrevo, which was later folded again into a division of Aventis. The point here is less the details than the big picture. There was lots of upheaval at Plant Genetics--its tiny U.S. headquarters moved through three cities in four years. It is reasonable to assume, too, that operational details surrounding a corn gene were hardly the most important concern of senior Aventis executives trying to manage a $20 billion merger. Until it was too late.

Although scientists at Plant Genetics were a few years behind the competition, they were excited about what they had created: a variety of the Bt protein that destroyed a different part of the corn borer's gut. This was important because an additional vulnerability would make it harder for the corn borer to develop resistance to Bt pesticides. The Bt variety created in Starlink corn was called Cry9. (Bt proteins have a crystalline shape, so different varieties were called Cry1, Cry2, etc.) Aventis scientists thought Cry9 was a winner that would make them significant players in the next generation of agricultural products. Federal regulators in the U.S. were more cautious.

In the early 1980s, when the prospect of bioengineered crops first emerged, people from numerous U.S. government agencies met to discuss how to regulate the products. They agreed that the Department of Agriculture would determine whether a new plant was safe to grow outdoors: Would it run amok, for example, and harm other plants or animals? If a genetically altered plant was supposed to produce a pesticide, the Environmental Protection Agency would decide whether the plant was safe in food. The Food and Drug Administration would enforce the food safety standards established by the EPA.

In 1997, when EPA scientists were evaluating Starlink, they saw something they hadn't seen in other brands of Bt corn. Starlink's Cry9 protein didn't dissolve in stomach acid as quickly as proteins in other Bt varieties. Nor did it break down as rapidly during cooking or processing. This meant that the Cry9 protein, unlike the others, might stay in the stomach long enough to be passed intact into the bloodstream, where it could trigger an allergic reaction. "Other Bt proteins lasted only a few seconds in simulated gastric juices," says Stephen Johnson, deputy assistant administrator of the EPA in charge of pesticide regulations. "This broke down much more slowly." In other tests, however, the Cry9 protein seemed fine. "We looked at the structure of the molecule and asked if it walked and talked like other known allergens," says Johnson. "It did not. So we were faced with two of three studies saying there was something different about this pesticide. We decided we couldn't allow it in food without more tests."

Starlink's developers, eager to market their product, invoked a little-known EPA rule that allows some pesticides and herbicides to be used on feed for animals but not on food destined for humans. This "split registration" had never been sought for genetically modified products, Johnson notes. "We looked at each other and said, 'What do we know about allergens? We know they don't pass through cattle.' We spoke to USDA and FDA, and they said [Starlink] passes the standard. We didn't feel real comfortable with it. But the law prevents us from saying, 'We don't like your product.' So we allowed it but put restrictions on it." For their caution, Johnson says, "we were denounced as pointy-headed regulators."

The restrictions on Starlink corn were severe. It could be grown only for animal feed or for nonfood use, such as conversion to ethanol. Because regulators worried that windblown pollen from Starlink stalks could pass the Cry9 gene to ordinary corn, farmers had to leave 660-foot buffer strips around their Starlink fields. Farmers bringing the corn to market had to notify grain elevators that it could not be used in human food. The EPA ordered Starlink's developers to require all farmers who bought the seed to sign a form affirming that they understood the restrictions and would abide by them. The company also promised to conduct a "statistically valid" survey of Starlink growers to ensure they were following the rules. Finally, says Johnson, "the company agreed to accept full liability if anything went wrong."

Neither Aventis nor its predecessor companies ever produced much Starlink corn. Instead they inserted the newly spliced genes into small amounts of corn and sold the resulting sprouts to seed companies. These then planted Starlink in greenhouses, harvested the corn, and replanted it to create more seed. Eventually the seed companies contracted with farmers who grow large volumes of corn for seed under controlled conditions outdoors. Once that seed was harvested, the companies had enough Starlink seed to begin marketing.

Ultimately, about a dozen small seed companies licensed Starlink corn from Plant Genetics. The Garst Seed Co., which is near Des Moines and has one of the longest pedigrees in the seed business, produced the vast majority of Starlink corn, according to Aventis executives. Garst, as is common with smaller seed companies, relies heavily on "farmer dealers" to sell its products. These are usually farmers who use the slow winter months to schmooze relatives and neighbors into buying a few thousand dollars' worth of seed. In 1998, the first year Starlink was on the market, just 10,000 acres were planted. Last year a mere 350,000 of America's 79.6 million acres of corn were Starlink. The highest concentration of Starlink in any state last year was 1.1% in Iowa, Garst's backyard.

Nevertheless, the proliferation of Bt corn was causing growing concern outside the Farmbelt. In April 1999 an entomology professor at Cornell University researching corn-borer resistance to Bt reported that he had fed a diet of corn pollen to monarch butterflies' larvae. Many of the monarchs that ate Bt pollen died. This caused a furor among environmentalists, who admire the monarch for its yearly migration from Mexico and back. Many environmentalists are profoundly worried about all genetically altered plants and animals, fearful that they contain health hazards that won't become apparent for years, or that they will somehow reproduce wildly and overwhelm ordinary species. For environmentalists, the monarch was about to become the poster butterfly of the anti-Frankenfood movement.

Among the environmentalists who led the charge against Bt corn was Larry Bohlen, an engineer by training and a senior official in the Washington office of Friends of the Earth. For years FOE and other greens had been trying to get the U.S. government to sign international protocols on the use of genetically modified organisms. "When the Cornell study on monarch butterflies came out, we had our first tangible example of the kind of impact genetic crops could have," says Bohlen. He wrote to President Clinton asking that use of Bt plants be suspended until their effect on nontarget animals could be determined. And he began writing to consumer-product companies like Campbell's, Kellogg, and Frito-Lay, urging them to forswear all genetically modified food. Last July the campaign began in earnest. Bohlen arranged for popular foods to be tested for genetically altered ingredients "so we could contact the manufacturers and tell them to be more careful."

Eventually Bohlen learned about Starlink. "When I asked grain-elevator operators and farmers how Starlink and other unapproved varieties were being segregated, I was told that separation was difficult and that very little segregation was being done." Bingo. Bohlen had his galvanizing image. "By summer it seemed there was a good chance Starlink had made it into the food supply." In late July of last year, Bohlen went to the Safeway near his home in Silver Spring, Md., and filled his grocery cart "with all the corn products I could find." He sent them to Genetic ID, an Iowa lab that routinely checks commodity shipments bound for Europe to make sure they comply with European Union standards. In September the news that Starlink corn had been found in tacos made by Kraft and sold under the Taco Bell brand was splashed across the front page of the Washington Post.

David Witherspoon, president of the Garst Seed Co., can't recall where he was when the news broke. That's surprising, because if anybody should have been electrified by the development, it was the head of Garst, which sold nearly all the Starlink produced in the U.S. "We were very concerned," Witherspoon now says. Aventis executives say they were flabbergasted and didn't believe the reports at first. A biotech industry organization immediately questioned the reliability of Genetic ID. But then Kraft ordered its own tests of the tacos; it found Starlink and recalled more than a million boxes. Other taco makers did the same. Kellogg shut down one of its mills because it feared Starlink contamination. Grain elevators, in the midst of gathering the fall harvest, scrambled for ways to test arriving truckloads for Starlink contamination. In many ways it was too late; most of the Starlink in the nation's food had come from the 1999 corn crop. And because 1999 had been a bumper year, there were more than a billion bushels of unsold corn still sitting in silos. No one knew how much of it was mixed with Starlink.

How did this happen? Every farmer who had bought Starlink signed a form agreeing to keep it out of the human food supply, right? Well, not exactly. Many of the 2,500 Starlink farmers appear to have been clueless about it. Hundreds claimed their seed salesmen never told them they were buying Starlink, and certainly didn't pass on any precautions about how to plant it. The head of the agriculture committee of the Iowa House of Representatives, Ralph Klemme, says he bought Starlink but was never told it was forbidden for use in food. Thomas Miller, the Iowa Attorney General, says "the vast majority" of farmers did not sign any forms acknowledging planting and marketing limits. It was not until a few weeks after the Starlink news broke that farmers who planted the seed received a letter asking them to sign and return some forms; the forms appear to have been backdated to before the spring planting. Aventis executives vigorously deny having anything to do with the letter. In a telephone interview, Garst CEO Witherspoon said he would "prefer not to get into that," citing potential litigation.

Witherspoon insists that Garst provided information to all its salesmen about Starlink. Asked whether Garst salesmen were diligent about having farmers sign the EPA-required forms, Witherspoon was vague. "The dealers would have started getting the forms and would know we had them. We tried to get them to dealers. We'd remind them to use them."

It seems unlikely that Garst's farmer salesmen would have knowingly deceived customers. The seed business relies heavily on the trust that exists when farmers sell seed to relatives and neighbors. Garst is one of the oldest companies in the business; it began in 1930 by marketing hybrid seeds developed by Henry A. Wallace, the founder of Pioneer Hi-Bred. (Wallace was later Vice President under Franklin Roosevelt.) Garst was so well known that Soviet Premier Nikita Krushchev visited its founder, Roswell Garst, on his Iowa farm in 1959. But the company ran into trouble in the early 1980s, when Pioneer severed its relationship with Garst to market its own seed. Garst lost the bitter lawsuit that ensued. The family sold the business to ICI, the British chemical company, in 1985. (Du Pont bought Pioneer in 1999.) ICI later spun off its U.S. seed business to Zeneca, a British drug company. Garst is now a part of Advanta, a joint venture between Zeneca and Royal VanderHave Group in the Netherlands. Ironically, Advanta made headlines in Europe last year when canola seeds it had sold there were found to contain small amounts of genetically altered material forbidden by the EU. The seeds, grown in Canada, may have been contaminated by windblown pollen from other canola nearby.

Aventis eventually took responsibility for the Starlink mess; the company is spending millions to locate the rogue corn so that it can be put into animal feed. Aventis executives say that they thought Garst was spelling out the restrictions on Starlink to farmers, but hint that they didn't monitor Garst carefully. Neil Harl, the agricultural economist at Iowa State, says he doubts Garst was motivated to be very explicit about how Starlink had to be grown and sold. "What farmer would buy a variety of seed if he was told he had to plant a 660-foot buffer strip around it, and would have to go through all sorts of special separation and storage after the harvest?" Witherspoon disagrees, saying the company sent 15 mailings to Starlink farmers. As for the "statistically significant" survey of farmer compliance that Aventis had promised the EPA, the company appears to have dropped the ball. Garst conducted the survey, says an Aventis executive, but did it right after the harvest, when most corn was still stored on farms.

Both Garst and Aventis officials implied in interviews that if they failed to live up to all their agreements with the EPA, it was because they were convinced Starlink would soon get full approval for use in food and that the special conditions would be lifted. "Aventis was working very hard on those approvals," says Witherspoon.

Even after giving Aventis and Garst their share of the blame, there's plenty more to go around. Johnson, the EPA official, now concedes that a split registration for Starlink, allowing it in feed but not food, was a dumb idea. "It was the first and last time we will allow that," he says. Critics point accusingly at the FDA, which was supposed to enforce food standards established by the EPA. Larry Bohlen at Friends of the Earth says the FDA didn't even have a way of testing for Starlink in food and that the agency moved slowly when news of the contamination first came out. "Kraft ran circles around the FDA. The day Kraft pulled its tacos off the shelf, the FDA was faxing me to ask if I would send them some of my taco shells. Kraft had already tested and confirmed on multiple lots." An FDA spokeswoman declined to comment on the agency's role in Starlink.

To its belated credit, Aventis has been aggressively trying to locate Starlink seed. It requested Garst's list of Starlink customers and met with all of them within days. Aventis is paying farmers up to 25 cents for each bushel of Starlink seed fed to animals. When grain-elevator owners discover that a batch of Starlink has contaminated a million-bushel silo, Aventis negotiates compensation for their added efforts and expense. The company has also paid for millions of test kits used by farmers, food processors, and grain handlers to identify traces of Starlink. Just how much is out there is anybody's guess. Because many farmers failed to plant buffer strips, pollen sometimes drifted into neighbors' fields, causing that corn to test positive. Moreover, some Garst seed varieties that weren't supposed to contain Starlink turn out to have been contaminated, the company now admits, and that adds to the difficulty of finding it.

Even though Aventis executives don't argue with assertions that the debacle may cost the company hundreds of millions of dollars, Wall Street appears unfazed. Aventis ADRs climbed from $71 to $77 between early September and late January. Some farmers have fared well too. Even if they planned all along to feed the Starlink they grew to their cattle, Aventis is paying them a premium for it.

But the reactions of people like Jerry Rowe are more typical. Rowe manages the Farmers Grain Cooperative, a four-million-bushel grain elevator in Dalton City, Ill. He says Starlink has greatly complicated his life. At peak times he unloads a truck every two minutes. "The Starlink test takes five minutes per truck, and I can't afford to slow down." And his sampling probe could miss Starlink lurking in a far corner of a truck. "Maybe I'll miss it coming in, but the customer finds it when I'm shipping it out," says Rowe. Corn that Rowe could sell for $2.14 a bushel to Archer Daniels Midland in nearby Decatur might get rejected, forcing him to spend 20 cents a bushel to ship it to Cedar Rapids, where the pay is just $2.06. Rowe also worries that if he finds Starlink in his bins a year from now, Aventis won't compensate him. Aventis claims that it will.

The long-term consequences of Starlink seed are hard to predict. No serious health problems have emerged so far. About three dozen people complained to the FDA about bad reactions to corn products in the days after Starlink first made headlines. Many clearly did not have allergic reactions, and virtually all the rest had mild problems like itchy eyes or a tight throat. A pediatric allergist from Duke University told a scientific advisory panel convened by the EPA that unless someone has an anaphylactic reaction to Starlink, he or she does not have a food allergy. But the panel decided that Starlink does indeed walk and talk like a potential allergen, and advised the EPA to turn down a request by Aventis that small amounts of it be allowed in the food supply.

Starlink has not triggered widespread hysteria about genetically modified food in the U.S., to the disappointment, no doubt, of some environmental groups. But Johnson at the EPA still worries that the episode may slow the acceptance of genetically modified products. "I am outraged at Aventis," he says. "This is enormously important technology. We trusted Aventis to handle it properly, and they didn't."

He is probably right to be concerned. Pierre Deloffre, head of a large French vegetable-processing company, told a seed trade convention in Chicago last December that Europeans turned abruptly away from genetically modified foods during the 1990s. Deloffre blames government regulators and scientists who failed to respond properly to Chernobyl, AIDS in the blood supply, and mad cow disease for eroding Europeans' confidence in technology. "Five years ago the first boatloads of genetically modified soybeans arrived here without the slightest reaction," says Deloffre. Now the EU barely touches them.

Although Europe hasn't imported much American corn for years, Japan is a large customer. The Japanese have been fairly tolerant of bioengineered food, but they, too, are growing cautious. New rules that take effect in Japan this spring will require labels on food to state if it contains genetically modified ingredients. An executive at ADM says orders from Japan for unmodified corn and soy have already begun to climb in anticipation of the new labels.

Since it caused no serious illnesses, Starlink will probably be a footnote in future agronomy textbooks. In reality, though, this was a disturbingly close brush with disaster. Starlink was probably circulating in the food supply for a year before it was found. If it had been slow acting but truly dangerous, like mad cow disease, the damage could have been enormous. Critical links in the food chain--from Aventis and Garst to thousands of small farmers--turned out to be either unconcerned about or oblivious to what they were selling and growing.

If we're lucky, maybe Starlink will also be a wake-up call, reminding us that tinkering with Mother Nature is risky business--and that it's not just white-coated lab technicians who must be careful. Solving the problem of hunger and malnutrition may ultimately depend not so much on science as on our faith in science and all its stewards. And if you can't trust a farmer, who can you trust?

FEEDBACK: boreilly@fortunemail.com