A congressional subcommittee upbraided the National Highway Traffic Safety Administration on Wednesday for failing to protect kids from injury and death in side-impact crashes and asked the agency to justify why it hadn’t adopted all of the tougher safety rules that members of Congress had pushed.

The scrutiny of car seat safety standards is part of a probe that the U.S. House Committee on Oversight and Reform’s Subcommittee on Economic and Consumer Policy launched this year in response to a ProPublica investigation. That story revealed that Evenflo, manufacturer of the popular Big Kid booster, marketed the seat as “side impact tested” when the company’s own tests showed a child using it could be paralyzed or killed in such a crash.

Evenflo was able to make up its own side-impact safety tests for boosters and assert that they passed them, because NHTSA never enacted side-impact test standards for children’s car seats and boosters despite a 2000 law directing it to do so. The bar was so low on Evenflo’s test, records show, that the only way its booster could fail was if the child-sized dummy was thrown onto the floor during a simulated side-impact crash or the booster broke into pieces.

Evenflo’s general counsel could not be reached for comment but has said in the past that the company has been a pioneer in side-impact testing and that its seats are safe, effective and affordable. Evenflo, a subsidiary of China-based Goodbaby International Holdings Ltd., has sold more than 18 million Big Kid boosters.

In a pointed letter to NHTSA on Wednesday, Rep. Raja Krishnamoorthi, the Illinois Democrat who chairs the subcommittee, and Rep. Katie Porter, a California Democrat, told NHTSA Acting Administrator James Owens that the subcommittee “is aware that manufacturers continue to take advantage of this key regulatory gap and market unsafe booster seats.”

In an interview, Krishnamoorthi questioned whether NHTSA, in allowing manufacturers to make up their own rules, has been captured by the industry.

“When we peeled the onion, we were just beside ourselves wondering what the heck is going on here,” he said.

A spokesman for NHTSA said that the agency had received the letter and looked forward to briefing the subcommittee “on the significant improvements in child passenger safety that NHTSA has made.”

This month, NHTSA proposed a new rule that would bar manufacturers from marketing booster seats to children weighing less than 40 pounds, a change that Krishnamoorthi and Porter had sought in March. As ProPublica’s investigation in February made clear, the agency for years has allowed manufacturers to label boosters as safe for children as light as 30 pounds even though the American Academy of Pediatrics has, for decades, said that those kids are safer in traditional car seats that use internal harnesses to hold their small bodies in place. Booster seats raise up children so that they can use vehicle seat belts designed for adults. ProPublica obtained internal testing videos and other documents that showed that Evenflo knew children weighing less than 40 pounds could be severely injured in side-impact crashes while seated in Big Kid boosters.

In their letter, Kristnamoorthi and Porter said they welcomed the higher weight requirement, but they stressed that they had also urged NHTSA to require prominent labels on boosters warning parents that they should only be used for children over 40 pounds. The letter criticized NHTSA for moving in the opposite direction in its new changes.

“Instead of adopting a clear labeling requirement, NHTSA proposed to ‘lessen restrictions on the labeling requirements,’ allowing manufacturers to present information ‘in their own words at locations that they deem most effective,’” they wrote, adding, “This will lead to confusion.”

NHTSA is known for moving at a glacial pace. The agency last proposed side-impact testing rules for children’s car seats in January 2014, but that proposal has languished for nearly seven years as manufacturers wrangled over what makes a good test. Car seats and boosters currently have to pass a test that mirrors the forces in a head-on collision. Even those tests are not up to date; they are done while the car seats and boosters are attached to a simulated back seat based on a 1974 Chevrolet Impala. This month, NHTSA also proposed a more modern back seat for the tests.

And the agency’s 2014 side-impact testing proposal doesn’t address a key finding of ProPublica’s investigation: It doesn’t apply to car seats or boosters for children over 40 pounds. In proposing to raise the minimum weight for boosters this month, NHTSA made it clear it plans to exclude boosters from side-impact tests it may later adopt. If those rules are enacted, booster manufacturers still would be able to make up their own tests and pass themselves.

Likewise, the standard NHTSA proposed in 2014 simulates a crash on the side of the vehicle nearest the child. While those crashes are dangerous, Krishnamoorthi and Porter urged NHTSA also to require a test that simulates when a vehicle strikes the side opposite where the child is sitting. Among side-impact collisions, NHTSA’s own data shows that those that occur on the side farthest from the child account for 40% of deaths and 30% of serious injuries for children seated in boosters and harnessed seats, Porter and Krishnamoorthi noted in their letter to NHTSA.

“The goal here isn’t just to say, ‘We, the government, did something,’” Porter said in an interview. “The goal is to keep kids safe.”

The danger of a far-side crash is that the body of a child seated in a booster can slip out of the vehicle’s shoulder belt and jacknife over the lap belt, resulting in spinal or head injuries.

The ProPublica investigation this year documented just such a crash on New York’s Long Island in 2016 that left Jillian Brown with an injury medical journals refer to as “internal decapitation.” Jillian, who at the time of the accident was 5 years old and weighed just under 37 pounds, was belted into a Big Kid booster that had a “Side Impact Tested” label stitched into the seatback. Jillian is now paralyzed from the neck down, steers her motorized wheelchair with her tongue and is kept alive by a ventilator. Sued by Jillian’s family, Evenflo blamed bad driving and said the seat performed as it was designed to do and didn’t cause her injury.

Court records filed this month show that Evenflo reached a confidential settlement with Jillian’s family.