DETROIT (AP) — Inside General Motors, they called it "the switch from hell."

The ignition switch on the steering column of the Chevrolet Cobalt and other small cars was so poorly designed that it easily slipped out of the run position, causing engines to stall. Engineers knew it; as early as 2004, a Cobalt stalled on a GM test track when the driver's knee grazed the key fob. By GM's admission, the defective switches caused over 50 crashes and at least 13 deaths.

Yet inside the auto giant, no one saw it as a safety problem. For 11 years.

A 315-page report by an outside attorney found that the severity of the switch problem was downplayed from the start. Even as dozens of drivers were losing control of their vehicles in terrifying crashes, GM engineers, safety investigators and lawyers considered the switches a "customer satisfaction" problem, incorrectly believing that people could still steer the cars even though the power steering went out when the engines stalled. In safety meetings, people gave what was known as the "GM nod," agreeing on a plan of action but doing nothing.

"The decision not to categorize the problem as a safety issue directly impacted the level of urgency with which the problem was addressed and the effort to resolve it," wrote Anton Valukas, the former federal prosecutor hired by GM to produce the report.

Some experts applauded the transparency in the GM report, but not everyone is buying its narrative, including family members of people killed and some lawyers suing the company.

Laura Christian, whose daughter Amber Marie Rose was killed in a Maryland Cobalt crash, still questions whether GM leaders knew about the problem — even though Valukas found that top executives, including CEO Mary Barra, didn't know about the switch problem until last December. Christian said the internal investigation is a start, but she hopes the Justice Department goes deeper and holds some employees criminally liable.

"Negligence is a criminal charge," she said.

The Valukas report makes no mention of negligence. But it says plenty about incompetence throughout GM.

THE NEW SWITCH

In the late 1990s, GM patented a new ignition switch designed to be cheaper, less prone to failure and less apt to catch fire than previous switches. But in prototype vehicles, the switch worked poorly. Veteran switch engineer Ray DeGiorgio had to redesign its electrical system.

The switch had mechanical problems, too. It didn't meet GM's specifications for the force required to rotate it. But increasing the force would have required more changes. So in 2002, DeGiorgio — who made several critical decisions in this case — approved the switch anyway. He signed an email to the switch supplier, "Ray (tired of the switch from hell) DeGiorgio."

Almost immediately, GM started getting complaints of unexpected stalling from drivers of the Saturn Ion, the first car equipped with the switch. The complaints continued when the switch was used for the Cobalt, which went on sale in 2004. Yet it wasn't seen as a safety issue. Even if the engine stalled and the power steering went out, engineers reasoned, drivers could still wrestle the cars to the side of the road.

As more complaints came in, GM kept viewing the problem as "annoying but not particularly problematic," Valukas wrote. "Once so defined, the switch problem received less attention, and efforts to fix it were impacted by cost considerations that would have been immaterial had the problem been properly categorized in the first instance," his report said.

In a critical failure to link cause and effect — and one that Valukas references often in his report — engineers trying to diagnose the problem didn't understand that the air bags wouldn't inflate in a crash if the engines stalled, failing to protect people when they needed it most.

In the meantime, GM customers, most unaware of the switch problem, kept buying the compact cars. Sales topped 200,000 in 2005, 2006 and 2007.

COMPANY INVESTIGATIONS

From 2004 to 2006, multiple GM committees with convoluted acronyms considered fixes without a sense of urgency, Valukas wrote. Crashes and deaths mounted, catching attention from company lawyers and engineers. Yet no one at GM figured out that the bad switches were disabling the air bags.

Fixes were rejected as too costly. Instead the company sent a bulletin to dealers explaining the problem and telling them to warn customers not to dangle too many objects from their key chains. GM elected not to use the word "stall" in the bulletin, saying that was a "hot" word that could indicate there was a more serious safety issue.

A Wisconsin State Patrol Trooper named Keith Young proved better at diagnosing the problem than GM employees, the report said. While investigating a 2006 Cobalt crash that killed two teen-age girls, he checked the wreckage and found the ignition switch in the "accessory" position; the air bags weren't deployed. Going further, Young found five complaints to government safety regulators about Cobalt engines stalling while being driven. Three drivers reported their legs touched the ignition or key chain before the engine quit.

Young also found the 2006 GM bulletin to dealers that detailed the switch problem. He determined that the Cobalt's ignition slipped into accessory before the crash, causing the air bag failure. A team from Indiana University that probed the crash in 2007 also made the connection. "Yet GM personnel did not," Valukas wrote.

They might have — if they read Young's report. An electronic copy was in GM's files in 2007, but no engineer investigating the switches reported seeing it until 2014, according to Valukas.

THE SECRET FIX

In 2007, John Sprague, an engineer working with GM's liability defense team, began tracking Cobalt air bag problems. He noticed a pattern and theorized a link to the ignitions. He also saw that the air bag problems stopped after model year 2007 and wondered if the ignition switch had been changed, Valukas wrote.

He was right, though he didn't know it at the time. In 2006, DeGiorgio had signed off on a change that increased the force needed to turn the key. But when asked in 2009 and later under oath, DeGiorgio denied making a change. "To this day, in informal interviews and under oath, DeGiorgio claims not to remember authorizing the change to the ignition switch or his decision at the same time not to change the switch's part number," Valukas wrote.

Keeping the same number prevented GM investigators from learning what happened for years, according to Valukas.

A 'BOMBSHELL' AND FINALLY A RECALL

By 2011, GM's outside lawyers were warning that the company could be facing costly verdicts for failing to fix the air bag problem. Company lawyers sought another investigation, but the engineer assigned to the case discounted the ignition switch theory.

The probe became stuck after two years with no results.

Then came what GM's outside lawyers called a "bombshell." An expert working for a law firm that was suing GM X-rayed two switches from separate model years and discovered they were different — GM's first knowledge of DeGiorio's change to the switch. Even so, GM's recall committee wasn't immediately told about the fatal accidents, so it waited for several months before it started recalling the cars in February, Valukas wrote.

Barra told GM employees Thursday that Valukas' report was thorough, tough and "deeply troubling." She said 15 people — including Ray DeGiorgio — were dismissed from the company and five others disciplined, and she outlined changes to make sure such a problem doesn't happen again.

But some have their doubts.

"If GM operated in the manner described over a full decade, then there are many more safety problems out there today," said Jere Beasley, an attorney who is suing GM on behalf of victims.