(SINGPORE 2026.8.25) If one day the workers for Myanmar’s scam compounds are no longer humans trafficked into forced labor, but AI agents that never sleep, never tire and can chat with hundreds of people all at once, would that sound like sci-fi?
It may not.
Recent reporting by QbitAI (量子位), a Chinese digital media outlet specialising in AI, highlighted a troubling development: AI is already being incorporated into scam operations in Southeast Asia, including the notorious scam compounds in Myawaddy, Myanmar.
Criminal groups are using large language models such as ChatGPT and Gemini to polish messages, translate languages, generate replies, and maintain conversations with potential victims. And these may be only the beginning steps.
A recent academic study has raised an even more uncomfortable possibility: AI may actually be better than humans at the most important part of a romance scam — winning someone’s trust.
The most effort-intensive part of a romance scam isn’t getting the money. It is getting someone to trust a stranger.
Scammers may spend days or weeks learning about their targets’ interests, remembering details about their families and jobs, responding to their emotions, and gradually creating a sense of intimacy. Only after trust has been established do victims begin to lower their guard.
More unsettlingly, the person making the victim fall in love may not be human at all.
According to an academic study highlighted by QbitAI, researchers conducted a week-long experiment comparing AI and human conversation partners.
They built a multi-module LLM agent that could simulate human typing and response delays, restart conversations after targets went silent, remember details discussed days earlier, and explain away its own mistakes.
Twenty-two participants were told that the study examined how people form relationships online. They were asked to chat with two people through WhatsApp for seven consecutive days — one human and one AI — without knowing which was which.
The results were eye-opening.
Forty-six percent of participants trusted the AI conversation partner enough to download an app it recommended, compared with 18% who trusted the human participant and downloaded the game that person recommended.
Even more tellingly, 80% of the participants’ messages sent were addressed to the AI. Participants gave the AI an average trust score of 3.78 out of five, compared with 3.31 for the human participant.
The findings suggest that AI may simply be better at performing the role of a sympathetic, attentive conversation partner.
Before participants were told the identities of their conversation partners, only one of the 22 participants correctly guessed that the other person was AI, according to the study.
The AI does not need to be perfect. It only needs to be convincing enough.
According to one report cited by QbitAI, investigators interviewed 145 people with inside knowledge of the scam industry. More than 100 were trafficking survivors who had previously been forced to work in scam compounds in Cambodia, Myanmar and Laos. Others included managers, money launderers, and AI specialists.
In 34 interviews conducted in 2024, every interviewee said they used ChatGPT in their daily work.
For now, AI is still mostly an assistant. Scammers use it to polish messages, translate languages, generate replies, produce opening messages in bulk, analyse previous conversations, and help human workers handle more conversations at the same time.
In other words, AI has not necessarily replaced the scammer. It has simply made the scammer more efficient.
But that could change.
C4ADS, a US-based security research organisation, investigated leaked records from scam compounds in Myawaddy and found evidence of sophisticated systems designed to industrialise online fraud.
That represents a fundamental shift: AI is moving from being an assistant to human scammers to potentially becoming the digital worker itself.
According to the investigation, one Indian man trafficked into the Tai Chang scam compound in Myanmar said he was expected to operate dozens of fake accounts and communicate with more than 100 people during a normal shift. In one month, he reportedly targeted approximately 50,000 people in at least 17 countries.
It was not a case of one scammer deceiving one victim. It was industrialisation of emotional manipulation.
According to US technology magazine WIRED, trafficked workers may still be more cost-effective now for scam organisations than paying for AI services.
People held in scam compounds can be forced to work for little or no pay, and some are reportedly forced to pay large ransoms before they can leave.
But AI could eventually change that calculation.
That would create a paradox: the number of people trafficked into scam compounds could fall while the number of scam victims rises.
Humans alone may not be able to fight AI-powered fraud on a massive scale. The answer will have to involve AI as well. China has already begun deploying AI-based anti-fraud and deepfake-detection systems.
According to material cited by QbitAI, China Mobile’s ( 中国移动 ) financial technology arm has developed technology designed to identify AI-generated face swaps within milliseconds, while social platforms such as Kuaishou (快手）have deployed AI models to intercept fraudulent activity.
Myanmar’s scam compounds are not simply a local criminal problem. They are a preview of what happens when organised crime, human trafficking, digital platforms, cryptocurrency, AI, and cross-border infrastructure converge.
Northern Myanmar was once a major centre of cyber fraud. After authorities intensified their crackdown, operations shifted, with Myawaddy emerging as another major hub.
In 2025, authorities targeted major compounds including KK Park and Asia Pacific New City, demolishing buildings and repatriating thousands of suspected Chinese nationals involved in online gambling and cyber fraud.
Yet the industry did not simply disappear. It moved. That is the important lesson.
AI could make that model even more mobile and resilient. If a physical compound is shut down, an AI-based operation could move parts of its infrastructure elsewhere with far less dependence on physical manpower.
Governments are therefore no longer simply fighting criminal organisations running buildings, workers, and bank accounts. They are fighting scattered digital systems that can operate across jurisdictions.
So, the most dangerous romance scam of the future may not begin with an obviously suspicious message. It may begin with someone who remembers your favourite food.
Someone who knows why you were upset yesterday.
Someone who asks about your parents.
Someone who sends you a good-morning message every day.
Someone who never seems too busy to listen.
Someone who appears to understand you better than anyone else.
Except that person may not exist at all.
He or she may be an AI agent — built, deployed and optimised by a criminal organisation to make you trust it.
