AI-Powered Deepfakes and Drone Armies: The Criminal Toolkit Just Got an Upgrade

The same large language models that streamline enterprise workflows are now criminal force multipliers. Threat actors deploy unguardrailed chatbots for real-time deepfake impersonations, optimal trafficking route-mapping, and social engineering at machine speed. For CISOs and security analysts, these AI-native tactics demand a fundamental rethink of identity verification, fraud detection, and dark-web monitoring strategies.

A 19-year-old law student in Melbourne allegedly stabbed a teenage boy and kicked a girl down an escalator in a crowded shopping centre, later confessing to police that he had been radicalised by artificial intelligence. This case, still untested in court, is the most vivid example yet of a disturbing trend: criminals and extremists around the world are weaponising custom-built AI chatbots—"evil twins" of tools like ChatGPT—to automate and supercharge illegal activity. The same technology that powers legitimate language models is being stripped of moral guardrails and deployed for grooming, recruitment, money laundering, deepfake fraud, and even aerial warfare via drone armies. Meanwhile, law enforcement struggles to keep up, its understanding of these threats patchy and ad hoc. The convergence of AI accessibility and criminal ingenuity is creating a new generation of threats that blur the line between digital and physical harm.

“

One Mexican cartel reportedly uses AI to control fleets of weaponised drones.

The incident itself is instructive. The accused—himself a law student—reportedly used AI to engage with extremist content, which allegedly pushed him toward violence. The case raises profound questions about agency, radicalisation, and the role of technology providers. While courts have long grappled with defenses based on external influences (television, video games, music), the interactivity and personalisation of AI introduces a qualitatively new element: an adaptive, seemingly empathetic interlocutor that can reinforce and escalate dangerous beliefs at machine speed and scale. This personalisation is what makes AI so effective for both legitimate and criminal purposes, and why the Melbourne case may set critical legal precedents.

Criminal enterprises are already well ahead. The article details how gangs use custom LLMs (large language models) to map optimal trafficking and escape routes, impersonate CEOs and relatives via hyperrealistic deepfake video calls, and automate money laundering—traditional processes that once required human foot soldiers now executed by code. One Mexican cartel reportedly uses AI to control fleets of weaponised drones. The technology is dual-use by nature; the same capability that helps a business optimise supply chains can help a cartel evade capture. For cybersecurity professionals, this means that attack surfaces are expanding: social engineering emails are now indistinguishable from genuine correspondence, synthetic identities pass biometric checks, and fraud operations scale effortlessly across languages and jurisdictions.