Marking the United Nations World Day Against Trafficking in Persons on July 30, observed this year under the theme “Trapped Behind the Scam” to highlight the trafficking networks that fuel cyber-enabled fraud, researchers from Amrita published a collaborative study with international partners from Ben-Gurion University, the University of Melbourne, Ca’ Foscari University of Venice, and Colorado State University.
Drawing on the testimonies of 115 survivors trafficked into scam compounds operating across Southeast Asia, the study warns that artificial intelligence is making scam operations faster, cheaper, more convincing and increasingly difficult to detect. The researchers note that these compounds are part of a highly organised transnational criminal enterprise in which victims, lured through fraudulent overseas job advertisements, are trafficked, confined and forced to carry out romance, cryptocurrency and investment scams under constant surveillance, intimidation and violence.
The study estimates that scams targeting people in East and Southeast Asia alone caused financial losses of between US$18 billion and US$37 billion in 2023, underscoring the scale of the rapidly evolving cybercrime ecosystem and the urgent need for coordinated global action.
One Scam, Two Victims
The person who sends you a romantic message or a good investment opportunity may not be an independent cybercriminal. They might be a trafficking victim being forced to reach out to strangers for hours a day.
Research shows that scam compounds operate like organized companies with managers, human-resource departments, training manuals, daily targets, and scripted conversations.
Trafficked workers may have their passports taken and be charged for food, accommodation and other basic necessities. Missing targets can lead to violence, threats or spiraling debt.
One survivor told researchers that workers were paid for “breathing seaside air.”
Much of this forced labour is spent cultivating friendships and emotional relationships over days or weeks before presenting a fraudulent financial opportunity. Researchers have found that artificial intelligence is increasingly helping, and perhaps automating, this process of building trust.
AI Is More Persuasive Than An Expert Human
Amrita’s top research project, “Love, Lies, and Language Models: Investigating AI’s Role in Romance-Baiting Scams,” will be featured at USENIX Security ’26, one of the leading global cybersecurity conferences.
The study found AI tools are already being used for translation, improving tone and drafting personalised responses to potential victims.
Participants believed they were interacting with two humans during the week-long experiment. In reality one of the conversation partners was an autonomous AI agent.
The AI complied with a request 46% of the time, compared with 18% for a trained human operator.
Commercial content filters also failed to detect the romance-baiting conversations because the messages initially appeared to be ordinary expressions of friendship, affection and emotional support. “Many of these scam messages are sent by trafficked people who are forced to send them. The same technology that scales the fraud can identify the fake job advertisement that lured them in. Current protections are insufficient and the burden is still on the targeted person. “This means we need more awareness and research,” said Dr Krishnashree Achuthan, Dean, PG Programs, Amrita Vishwa Vidyapeetham.
Spotting Fake Job Offers Before Trafficking Starts
The researchers also have developed a machine-learning system capable of identifying fake job ads used to recruit trafficking victims.
The system was trained on a multilingual dataset of job postings in English, Hindi and Bengali to identify warning signs like unrealistic salaries, vague job descriptions, urgent hiring requests and suspicious overseas employment offers.
The technology could eventually be built into job portals and browser extensions to alert applicants about potentially dangerous ads, the researchers said.
