There might be a huge flaw in the government’s system to protect you from salmonella.

One of the US Department of Agriculture’s main tasks is to ensure that the nation’s meat supply is safe. But according to a new peer-reviewed study from the department’s own researchers, the USDA’s process for monitoring salmonella contamination on chicken—by far the most-consumed US meat—may be flawed.

The process works like this: After birds are slaughtered, plucked, and eviscerated, the carcasses are sprayed with a variety of antimicrobial chemicals designed to kill pathogens like salmonella and campylobacter, and then plunged into a cold bath (which also includes antimicrobial chemicals) to lower their temperature. At that point, a few of the birds are randomly selected, rinsed, removed from the line, and put into plastic bags filled with a liquid that collects any remaining pathogens. The liquid is then sent to a lab for testing within 24 hours. (The test birds go back into the production line.) If a large number of them test positive for salmonella, the USDA knows there’s a problem and takes steps to address it.

According to the USDA’s Food Safety and Inspection Service (FSIS), which oversees safety protocols in the nation’s slaughterhouses, the salmonella system works great. The agency’s latest numbers show a steadily falling incidence of positive tests for salmonella on chicken carcasses: just 3.9 percent in 2013, down from 7.2 percent in 2009.

But a new study by scientists at the USDA’s Agricultural Research Service (ARS) paints a less rosy picture. The researchers simulated the FSIS’s method for testing collecting pathogens from chicken carcasses, and found it can turn up negative results even when salmonella is present.

Here’s why: When those birds are plucked off the line for testing, they’ve just been bombarded with antimicrobial chemicals, and traces of those chemicals can collect in the testing bags along with remaining microbes. In order for tests to be accurate, the germ-killing chemicals have to be quickly neutralized by the testing liquid. If they’re not, they can keep killing bacteria and, as the study puts it, “lead to false-negative results due to sanitizer carryover into the carcass.” And that’s exactly what happened in the simulation the researchers conducted. The authors concluded that their study “suggests that current procedures for the isolation and identification of Salmonella on poultry carcasses may need modification.”

But the FSIS disagrees with this conclusion. “FSIS is confident that our testing results yield accurate outcomes,” an agency spokesman wrote in an email. He emphasized that the ARS study was a simulation, and “did not evaluate the same practices as our in-plant personnel utilize.”

Salmonella poisoning remains a huge problem. Starting in March 2013, a salmonella outbreak traced back to chicken sickened more than 600 people in 29 states, 38 percent of whom had to be hospitalized. And—unlike the FSIS’s tests for salmonella on chicken carcasses—salmonella poisoning rates have not shown any steady decline pattern over the past 15 years