This is a research prototype. The data and analyses are preliminary and not yet validated — we'd welcome your .
Traffickers in Myanmar are using AI-powered software platforms, built with models from OpenAI and Google, to facilitate industrial-scale romance and investment scams. These tools allow forced laborers to manage hundreds of victim profiles simultaneously across dozens of languages. The investigation highlights how AI infrastructure is being weaponized to maximize fraud efficiency while simultaneously surveilling and punishing the trafficked workers operating the systems.
AP reported that traffickers at Myanmar's Tai Chang scam compound forced a man named Safeer Mohammed Koorimannil to use the AI-powered KT platform to impersonate women and target thousands of people online. The system reportedly supported the scam conversations while monitoring worker performance. Koorimannil said he was beaten for poor results and later paid 500,000 rupees ($5,300) for his release.
Domain classification, causal taxonomy, severity scores, and national security assessments were LLM-classified and may contain errors.
Using AI systems to gain a personal advantage over others such as through cheating, fraud, scams, blackmail or targeted manipulation of beliefs or behavior. Examples include AI-facilitated plagiarism for research or education, impersonating a trusted or fake individual for illegitimate financial benefit, or creating humiliating or sexual imagery.
Also classified as
Victims of the romance scams developed deep emotional trust and reliance on the AI-generated personas, leading to severe financial and psychological exploitation.
The primary use of the AI systems is to conduct highly targeted romance and investment scams to defraud specific victims for financial gain.
Human
Due to a decision or action made by humans
Intentional
Due to an expected outcome from pursuing a goal
Post-deployment
Occurring after the AI model has been trained and deployed
The risk is caused by human scammers and traffickers intentionally deploying and abusing AI models to automate fraud and exploit workers after the models have been deployed.
Added categories peak at 3: Substantial (Psychological Harm, Indirect)
Transnational criminal networks in Myanmar are leveraging advanced U.S. AI models to scale global financial fraud and romance scams. This AI-enabled industrialization of crime relies on severe human rights abuses, including the forced labor of hundreds of thousands of trafficked workers, posing substantial challenges to global economic security and human rights.
The AI models ChatGPT and Gemini functioned as assistants to the scammers, generating automated replies, translating messages, and powering role-play chatbots to help scammers draft convincing messages, requiring human scammers to copy, paste, and send them.
Unacceptable Risk: The report describes AI systems being used for subliminal manipulation and behavioral influence to defraud victims of their life savings, which poses significant dangers to society and violates fundamental rights.
The AI models were primarily used to power role-play chatbots and generate automated conversational replies to scam victims, as well as translate messages across dozens of languages.
The report is a highly detailed, collaborative investigation by AP and FRONTLINE, utilizing leaked files, blockchain analysis, device data, satellite imagery, and interviews with 58 victims and 36 scammers. The evidence linking AI models to the scam operations is robust and explicit.
None.
Technology
AI Model Exploitation
ChatGPT and Gemini are integrated into gray-market scam software.
Automated Translation & Chatbots
AI powers role-play chatbots and translates replies across 100+ languages.
Evasion of Platform Safety Checks
Scammers route traffic through US cloud services to bypass checks.
Data Inputs
Scam Script Training Data
Scam bosses generate scripts used by AI to generate automated replies.
Valuable Leased IP Addresses
Scammers lease US IP addresses to make traffic appear trustworthy.
Human Factors
Trafficked Worker Coercion
Forced laborers operate AI-powered tools under threat of physical violence.
Victim Psychological Manipulation
AI-generated intimacy exploits vulnerable victims in romance scams.
Process and Methods
Reactive Abuse Mitigation
Tech companies rely on post-facto account bans rather than prevention.
Gray Market Software Distribution
Commercial software like KT and 007TG is sold openly to scammers.
Evasion of Telecom Blockades
Scammers migrate to Starlink and new compounds to evade local crackdowns.
Management
Prioritizing Revenue Over Risk
ISP and tech revenue from routing addresses outweighs the risk of action.
Inadequate Service Monitoring
Providers cite privacy by design to avoid proactive monitoring of traffic.
Failure to Enforce Terms
Tech companies do not vigorously enforce terms prohibiting fraud.
Regulatory Environment
Lack of Legal Incentives
US tech companies lack legal and financial liabilities for facilitated scams.
Voluntary US Cooperation
US regulators request voluntary cooperation instead of enforcing penalties.
Inconsistent Global Regulation
Strict laws exist in the UK/EU/Singapore but not in the United States.
AI system involved
Identified by automated extraction; can differ from the alleged developers above.
