Every major shift in communications technology has reshaped trafficking — classified ads, social media, encrypted messaging, crypto. Generative AI is the latest, and it is cutting both ways faster than any of its predecessors. Criminal networks are using it to scale deception; investigators are using it to make sense of data volumes no human team could read.
The question is no longer whether AI belongs in the fight against trafficking. It is who uses it better, and under what rules.
The offense: exploitation at lower cost
A 2026 webinar series run by the Bali Process Regional Support Office and the OSCE concluded that operations which once needed significant staff can now be run with free or low-cost AI tools. Deepfakes, jailbroken language models and AI-assisted malware make identity theft, extortion and large-scale fraud cheaper to run and harder to detect — and law enforcement has limited visibility into how much of this is already happening.
UNODC’s July 2026 assessment of Southeast Asian organized crime documented syndicates using generative AI, deepfakes and automated fraud. In the region’s scam compounds, trafficked workers and AI tooling operate together: the tools multiply the reach of every coerced worker, and the coerced workers supply the human judgment the tools still lack.
Sextortion: where AI harm reaches children
Financially motivated sextortion — coercing victims into sending intimate images, then demanding money — is one of the fastest-growing online harms. FinCEN’s September 2025 notice to banks reported nearly 55,000 related FBI complaints in 2024, a 59% annual increase, and warned that boys aged 14 to 17 are increasingly targeted. NCMEC has linked financial sextortion to more than two dozen suicides in the U.S. and Canada.
Generative AI removes the need for a victim to share anything at all. Offenders can take an ordinary photo from social media and fabricate explicit imagery. Research by Thorn with NCMEC found that roughly one in ten financial sextortion reports in a single year involved inauthentic images. FinCEN has asked financial institutions to flag payment patterns consistent with these schemes, turning banks into an early-warning layer.
For victims and families, NCMEC’s Take It Down service and CyberTipline remain the first points of contact.
The defense: finding signal in the noise
Trafficking investigations generate data that is huge, messy and deliberately obscured: thousands of ads, rotating phone numbers, coded language, fragmented across platforms. AI is well suited to that problem.
The best-documented example is DIG (Domain-specific Insight Graphs), built at the University of Southern California’s Information Sciences Institute under DARPA’s Memex program and described in a 2026 AAAI paper. DIG ingests web-scale advertising data, normalizes it and links entities — phone numbers, emails, locations — into a knowledge graph that non-technical investigators can query. It was used by more than 200 U.S. law-enforcement agencies for over five years. In New York, agencies using it reported trafficking-related arrests rising from under 1% to over 60% of relevant arrests while arrests of sex workers fell — the outcome advocates have long sought: targeting exploiters rather than the exploited.
New tools are coming through. In July 2026 UN Human Rights and Call for Code AI announced three winning applications — TEND, Sentinel AI and TRACE — from a global challenge launched on the centenary of the 1926 Slavery Convention. Judges were drawn from law enforcement, survivor services, human-rights law and AI engineering. The challenge’s premise was to turn the same platforms traffickers use to recruit and advertise into sources of detection.
Guardrails that make the difference
Investigative AI can fail in ways that harm the people it is meant to protect: false positives that criminalize sex workers or migrants, surveillance that sweeps up survivors, or biased models that over-police particular communities. The emerging consensus across prosecutors’ networks, the OSCE and academic legal research is consistent:
- Augment, don’t replace. AI should surface leads; trained investigators and survivor-informed services should make decisions.
- Explainability and admissibility. DIG’s developers stress that usability, explainability and policy compliance were as important as the algorithms — evidence has to stand up in court.
- Rights and data protection. Tools must comply with human-rights and privacy law, especially facial recognition and cross-border data sharing.
- Detect, investigate, collaborate. The Bali Process/OSCE framework calls for screening all evidence for AI-generated content, building forensic capacity and sharing tactics across borders and with industry.
- Safety by design. Pressure is building on AI developers to build abuse safeguards into generative models and chatbots before deployment.
- Sustainability. DIG’s history shows a recurring risk: tools proven in pilots can stall when research funding ends.
What to watch
- Legislation on AI-generated child sexual abuse material and non-consensual deepfakes across U.S. states and the EU.
- Whether the 2026 UN challenge tools move from prototype to operational deployment with funding behind them.
- Financial-intelligence typologies that link sextortion payments, scam-compound revenue and trafficking networks.
- Industry safeguards in generative-AI products, especially image tools and conversational agents.