When Tank, a West Valley City police dog, began sniffing around the silver Mazda Protege during a traffic stop last February, the police officers with him had a hunch that there was drugs in the car.

Detective David Allen had been watching their suspect, who he thought he’d seen buying drugs weeks ago. Allen pulled him over for speeding, and called Officer CJ Moore to bring Tank.

During the search, Tank sniffed the car, but was sometimes distracted by passing traffic or something on the sidewalk. He never “alerted” like he was trained to do when he caught the scent of narcotics, but Moore told his fellow officer that he knew from his dog’s “natural behaviors” that the animal had smelled drugs.

So the officers searched the car, and found a marijuana pipe with some residue and a gun that the man wasn’t supposed to have because he was a felon. He was charged in federal court with two crimes.

But in a ruling handed down last week, U.S. District Judge Clark Waddoups threw out the charges and questioned whether Tank and other police dogs are being properly trained by Utah officials before being sent out to help police catch drug dealers.

Scott Stephenson, the director of Utah’s Police Officers Standards and Training (POST), said Friday that they’re looking at Waddoups’ ruling and deciding whether their K9 training program needs to change.

“We’re taking it to heart,” he said, “and evaluating current practices.”

Waddoups sided with a defense expert who testified at an evidentiary hearing that Tank’s actions of sniffing the wind, pinning his ears back and moving his head back and forth quickly is not a sign that he smelled drugs — it’s just what any dog would do when sniffing something.

The expert, who trained police dogs for more than two decades, testified that she watched body camera video of the traffic stop and could not pinpoint any moment where Tank showed he had smelled drugs. Waddoups said he couldn’t either in reviewing the footage several times.

“A K9’s trained final response is an objective and ‘definitive signal’ that a dog has detected the odor of a narcotic in the vehicle,” Waddoups wrote, “but here Tank did not display his final trained response. Rather, the determination that Tank detected narcotics was made by Officer Moore and was based on his interpretation of Tank’s behavior and his subjective classification of those actions as ‘alerts.’”

The federal judge also wrote that he had “serious concerns” about whether POST properly trained Tank. He noted that when police dogs are certified, they should go through “double-blind training,” where neither the handler nor the person conducting the test knows during training if there are drugs present or where they are hidden. The defense expert testified that dogs are sensitive to people’s body positioning and facial expressions, which can inadvertently impart bias on the animal — leading them to “indicate” they had found something when they haven’t.

Waddoups noted that POST’s training and tests cast doubt as to whether the dogs could be trusted to alert when they actually smell drugs, or if they are just trying to please their masters.

“The final test that a K9 must pass in order to be certified is not even performed single-blind,” Waddoups wrote. “As such, the K9’s handler in the exam, who is the same officer who has worked with the K9 for months and has a clear interest in having his K9 be certified, knows exactly how many [drug] hides will be present in the exam and can therefore continue to search until the K9 finds them all.”

Stephenson pushed back on that assertion, saying the handlers don’t know how many hides there are — only the person judging the dog does. The director said that other states also rely on single-blind testing, and that’s how it’s always been done in Utah’s program.

“We’ve had several audits done on our program,” he said. “In fact, two recently last year that say that single-blind testing is effective. But when a local judge rules, you have to pay heed to that.”

West Valley City police officials did not respond to a request for comment.

Waddoups wrote he was also concerned that Tank only had gone through four narcotics training after he was certified, almost always in scenarios that would have made it impossible for him to make a “false-identification of narcotics.”

The federal judge said that because Tank did not clearly indicate that he had found drugs in that Mazda Protege, the police officers had searched the man’s car without probable cause to believe he had committed any crimes. He noted that the police officers already believed there would be drugs in the car, and that could have inadvertently influenced Tank’s behavior.

“A K9 sniff cannot simply be a formality or an excuse to support a search,” Waddoups wrote. “... in the heat of the hunt, judgment can be compromised and even zeal to apprehend a perceived criminal may influence a decision that would not otherwise be justified.”

After a March evidentiary hearing, Waddoups had ordered both sides to file written legal arguments about whether Tank’s training was sufficient and if there was probable cause for the officers to search the man’s car.

Instead, federal attorneys asked for the charges to be dismissed because of a “lack of prosecutable evidence and for good cause.” Waddoups granted the dismissal, but also ruled on the motion to suppress the evidence that police found in the man’s car — saying that the conduct of officers should be looked at to “provide guidance in the arena of K9 sniffs.”