In the past two years, the North Side-based company has used 60,000 posters to identify 734 trafficking victims, all 13 to 25 years old, almost entirely women, and 84% people of colour. — Photo by Steffen Schubert on Unsplash
Human trafficking networks are complex.
But for more than a decade, artificial intelligence has helped one Pittsburgh startup unravel the complexity to spot trends that can help rescue victims and even prosecute those responsible for their exploitation.
To explain how it works, Marinus Analytics co-founder Cara Jones pulls up a website on her phone.
This particular site is free, and hosted in Europe, but within seconds it connects her to a woman in Western Pennsylvania. Users can sort by location and specific sexual interests. There are photos and descriptors that feel reminiscent of an Amazon listing.
Each label is a red flag to Jones, who has more than a decade of experience working with the Department of Homeland Security and other agencies to catch those preying on vulnerable young women.
"New in the area" means this woman was potentially brought to Pittsburgh by people profiting from her work, she explains. "Exotic Latina" could imply the involvement of foreign nationals. "24/7" availability suggests a lack of boundaries that independent sex workers would typically establish.
This was a random person Jones found within seconds, and already she has identified three "alarming" signals. The next step would be to check her face against Marinus' catalog of missing people, which is refreshed daily from law enforcement postings across the country.
In the past two years, the North Side-based company has used 60,000 posters to identify 734 trafficking victims, all 13 to 25 years old, almost entirely women, and 84% people of colour. By flagging these cases, Marinus is able to trigger investigations by federal agencies that can help free the women and prosecute trafficking as a federal felony.
The crime is more pervasive and nuanced than people may realise: It doesn't require victims to be moved across state lines, as the 1910 Mann Act once established. To be considered trafficking, prosecutors must prove that force, fraud or coercion were used to compel someone to provide labor or services, including commercial sex. It's not an easy legal standard and often requires strong regional cooperation between agencies to investigate, Jones said.
On Dec 10, she was called to testify before the House Subcommittee on Cybersecurity, Information Technology and Government Innovation on how AI can combat trafficking.
Chairwoman Nancy Mace and other lawmakers described the crime as heinous but rarely discussed and said it's time to use technology to fight back.
"Artificial intelligence, data analytics and digital forensics are no longer science fiction," she said. "They are recovering victims in hours instead of months, identifying networks once hidden in plain site and building courtroom-ready cases against predators who thought the internet made them untouchable."
Mace, R-S.C., held a similar hearing on how to use modern tools to combat illegal immigration earlier this year.
Marinus' software use is narrow and the victims clearly defined, Jones said. The system is used by about 250 law enforcement groups around the world. It's overseen by fewer than 20 employees.
Colin Ward, now Marinus' head of global engagement, first started using the software as a detective in the United Kingdom. His best alternative at the time was going to Google and doing unstructured searches.
"You're missing lots of things" with that approach, and it's impossible to be proactive, he said.
Marinus' artificial intelligence models parse vast swaths of public data and use tools like facial recognition to string important information together, saving time and lowering the administrative burden of investigations, ultimately making it easier to rescue victims, company officials said.
The tool helped law enforcement find people who were advertising a 14-year-old girl for sex, Ward said. It also helped rescue a woman who had been kidnapped. Even though her face was covered with tape, the software was able to identify a tattoo on her neck.
Ward describes Marinus as a simple solution to a complex problem, though he said it takes a team of "very clever" technicians to make the AI hum seamlessly.
These humans also serve as a crucial check on the technology's conclusions, ensuring they don't accidentally pursue a false target. On the flip side, the data can help "back up a hunch" that human investigators already have, Ward said.
As the company expands its global footprint, he is meeting with leaders from various countries, including a recent visit from Brazil, to establish relationships. One thing they want to ensure is that officials don't exploit the technology to target sex workers.
Ward sees the model built in Pittsburgh as one that could be replicated around the world. And he said he's not chasing that goal for profit.
"I'm coming from a policing background where I dealt with victims," he said. "All I want is for other countries to use that tool to safeguard victims." – Pittsburgh Post-Gazette/Tribune News Service
