The technology could allow scam syndicates to carry out multiple tasks to cheat victims at the same time without needing human intervention
Online scam syndicates that rely on trafficked workers, guarded compounds and vast call-centre-style operations could soon become more threatening by exploiting agentic AI to run some fraud operations autonomously, a cybercrime expert has warned.
Unlike earlier uses of artificial intelligence to write scripts, translate messages or produce deepfakes, agentic AI systems can carry out multi-step tasks with limited human prompting, raising the prospect of scams that are faster, cheaper, more personalised and much harder for authorities to trace.
scamsin Asia-Pacific tripled to US$114 billion last year from 2023, according to a United Nations report released last month.
The surge was due to an “increasingly integrated, transnational, and technologically sophisticated criminal economy” that shared expertise in money laundering, human trafficking and data harvesting, the report said.
Jonno Newman, a cybercrime expert with on-chain intelligence platform TRM, said the complexity of scam operations in the region could grow exponentially once criminals figured out how to use agentic AI effectively.
Newman said syndicates were already investing heavily in training agentic AI agents to become tools capable of harnessing years of data compiled by their large language model predecessors to target multiple victims across different languages.
“So instead of one human, they can use a bot that can do the work of 10 people. You would no longer need a big fortress – a scam operation could be set up from anywhere remotely,” Newman told This Week in Asia.