When a federal agency sought a man wanted in a firearms-related incident two years ago, it hit roadblocks. The informant knew no names—only that the criminal's sister, who lived in Tucson, Arizona, had an abusive boyfriend.

In a series of searches, officials combed hundreds of case files until they found a woman involved in a domestic-violence case who was also linked to a man of similar age with the same last name. And they nabbed the criminal. This should have taken at least a month. It took 25 minutes. It was one of Coplink's earliest successes.

Coplink, an artificial-intelligence–driven search engine for crime characteristics, scans multiple databases for connections among names, vehicles, physical descriptions, and other aspects of a crime or criminal. Developed by Hsinchun Chen, director of the University of Arizona Artificial Intelligence Laboratory, Coplink began in 1997.

Five years later, Chen has deployed Coplink at six agencies and is developing an information-sharing and analysis program with the CIA. He's also talking with the FBI about Coplink's potential use in counterterrorism efforts with the FBI. Coplink also helped build the case against the two sniper suspects in the Washington-area string of shootings this past October.

Coplink differs from standard database software in that it uses an AI component. A neural network continually updates information from many databases. It learns patterns of association that help it perform multiple searches with increasing intelligence.

The system is available only to authorized law enforcement personnel at www.coplinkconnect.com. Chen says that Coplink's security is strong enough to ward off hackers. "It's ten times more secure than most of the e-commerce environment," he says. "It's a supercop." Without the doughnuts.