NAGOYA - Around 6:30 a.m. on May 14, a detective was walking through Kanayama Station in Nagoya, central Japan, on his way to work. As one of the city's major transportation hubs, it is bustling with people from early in the morning.
Suddenly he spotted a face in the crowd that resembled a person who had long been a thorn in the side of the Handa police department.
Since the start of 2026, there have been 10 incidents of car break-ins in the jurisdiction. The targets were apparently anglers who were away from their vehicles for long periods of time. The crimes had become a major police concern.
Investigating police obtained security camera footage showing a person believed to be the suspect. However, the face was blurry, and they had been unable to identify the person.
But the man the police officer spotted at Kanayama Station looked like the person in the footage.
The detective made a split-second decision to follow him. The man boarded a Meitetsu train heading toward the Minamichita area, which is known for its many fishing spots. While contacting a police colleague who commutes through the same station to request backup, the detective boarded the same car as the man.
After getting off the train, the man headed toward the beach. The detective had tracked him for a total of three hours, exercising utmost caution but increasingly certain it was the suspect. When the man smashed the window of a car in a parking lot for anglers and searched the interior, the detective arrested him on the spot.
The Handa police department subsequently continued its investigation and, on June 11, rearrested the man on suspicion of fraud for allegedly using a credit card stolen from a vehicle in one of the 10 cases.
Despite the advent of artificial intelligence in law enforcement, Japanese police continue to use old-school investigative techniques such as visual identification searches. This involves officers memorizing the facial photographs of numerous suspects and standing at train stations and in busy downtown areas to spot them. They also watch for the characteristic eye movements of pickpockets as they scout for targets.
On the other hand, with recent advances in technology such as security cameras, relay investigations, in which investigators track a suspect's movements by following footage from security cameras along their escape route, have proven highly effective.
Police are also using AI to search video footage for people or vehicles of interest.
One of AI's strengths is that it simplifies the process of manually reviewing the large volume of video footage collected through relay investigations to identify suspects, allowing them to be located quickly. It also reduces the workload on investigators.
In an era where AI is accelerating technological progress, some police officers question whether traditional methods are still useful.
But there are some methods in which human skill is recognized as superior to technology. A prime example is the police sketch of a suspect, drawn by forensic investigators based on descriptions of the suspect's features provided by witnesses.
The Aichi Prefectural Police also holds annual training sessions where young police officers learn how to draw these sketches.
In the past, police used composite sketches. They are created by having witnesses select the features that most closely resemble a suspect's eyes, nose, and other facial features, and then combining them in a montage.
The technique was used in the "300 Million Yen Incident," one of Japan's most notorious unsolved heists. On Dec. 10, 1968, in Fuchu, Tokyo, a young man posing as a motorcycle police officer stopped bank employees transferring 300 million yen (about $833,000 in 1968) in bonuses, falsely claiming that their car was rigged with dynamite to explode. Once they exited the vehicle, he drove off with the cash and was never identified.
Montage photos of the suspect required specialized equipment and took a long time to produce.
As a result, police sketches, drawn on-site by investigators in a short amount of time while witnesses' memories are still fresh, have become the standard. There have been cases where suspects were identified and arrested based on police sketches. Even now that AI can produce lifelike images, the practice has not fallen out of favor.
"Even if a sketch isn't an exact likeness overall, one that emphasizes the details and features that stuck in a witness's mind is more likely to lead to an arrest," said a veteran investigator who conducts analyses of sketches.
In criminal trials, however, objective evidence, such as fingerprints and video footage, is required to prove guilt. In the case of the Handa vehicle break-ins, the thief was caught red-handed.
"We can't make an arrest based solely on an investigator's testimony, but in this case, it would have been difficult to do so even with just the video footage. It was a team effort, including support from other units," said a senior official at the Handa police department.
He stressed that both investigative experience and digital technology were crucial.
"We're sorry that damage was caused. However, considering the need to secure a guilty verdict and an appropriate sentence, arresting someone in the act -- when we can observe the crime firsthand and obtain conclusive evidence -- is an effective approach."
The work of a detective requires the ability to respond flexibly to situations that change from moment to moment and to make sound judgments. Even if you ask AI, it probably would not be able to provide an effective answer.
