Study on the recruitment of organised crime through AI

The EL PACCTO 2.0 Innovation Lab examined how criminal organisations use artificial intelligence to recruit children, adolescents, and adults across Latin America, the Caribbean, and the EU.

Six takeaways:

1. Recruitment is the entry point to nearly every form of organised crime — and AI is reshaping it. AI doesn't change the underlying logic; it lowers the cost, skill, and language barriers while increasing the speed, scale, and precision with which networks reach, profile, and deceive targets. The result is a more professional, industrialised recruitment process.

2. Violence-as-a-Service (VaaS) is scaling online. Networks now outsource violence — from intimidation to contract killings — to third parties recruited through social media, gaming, and encrypted apps. Europol's Operational Taskforce GRIMM identified 1,400+ individuals linked to VaaS networks and made ~280 arrests in its first year. Minors, some as young as 13–14, are recruited by design, for their reduced legal exposure.

3. This is a cross-regional threat. Documented cases span the LAC–EU corridor: scam centres (including inside prisons in Colombia and Bolivia), cartel and gang recruitment in Mexico and Ecuador, VaaS networks operating across Sweden, Denmark, and Norway, and trafficking flows moving victims from Latin America and Eastern Europe into Western Europe.

4. The tools are mostly commercial and dual-use. Mainstream LLMs, AI translation suites, deepfakes, and voice cloning do most of the work — alongside uncensored open-source models and a growing "Crime-as-a-Service" market (WormGPT, Xanthorox, and 200+ malicious variants) that removes the skill barrier for less capable actors.

5. The trend points one way: automation. Fully AI-powered and agentic recruitment — targeting, engaging, and building trust with no human in the loop — is nascent but already credible. AI is also generating novel tactics that don't match the patterns law enforcement is trained to recognise. The threat being monitored may not be the threat that emerges.

6. The vulnerable are targeted because of their vulnerability, not despite it. AI-generated recruitment content is often indistinguishable from legitimate material and slips past content moderation. Behind every case is a person: a child groomed into exploitation, an adult deceived into forced labour, someone coerced into a criminal role.

The full study sets out nine recommendations for law enforcement, platforms, and policymakers.