US AI Models Power Myanmar Scam Networks at Industrial Scale

Trafficked workers use American AI to run romance scams at scale

A trafficked worker in Myanmar used ChatGPT and Google's Gemini to simultaneously impersonate dozens of people and target 50,000 victims across 17 countries in a single month, according to an investigation by The Associated Press and "FRONTLINE."

Safeer Mohammed Koorimannil, who was forced to work at a scam compound in Myanmar, chatted with more than 100 victims at once across multiple fake profiles while supervisors patrolled with electric batons. His targets included a widowed tailor in Kurdistan, a sheep farmer in Kyrgyzstan, soldiers in Iraq, and dozens of others—all contacted through specialized software built on American AI models.

The investigation, based on leaked scam center files and analysis of over 200,000 device connections, reveals how US technology companies inadvertently enable industrial-scale fraud that cost Americans nearly $200 billion in 2024, according to Federal Trade Commission estimates.

Why it matters

The scam industry's exploitation of American AI and internet infrastructure represents a fundamental shift from individual con artists to industrialized fraud operations. Without stronger regulatory incentives, US companies lack compelling business reasons to invest in preventing abuse of their platforms—even as other countries including the UK, EU, Australia, and Singapore introduce mandatory anti-scam requirements with financial penalties.

AI-powered tools eliminate language barriers

Scammers purchase specialized software platforms like Kongtian Intelligent Customer Acquisition (KT) and Global Social Traffic Navigation (007TG) that integrate ChatGPT and Gemini for real-time translation across 100+ languages, automated response generation, and worker performance tracking.

Blockchain analysis by TRM Labs found that a single cryptocurrency wallet used by 007TG received $860,000 in payments between April 2024 and December 2025. The scammers who purchased these tools generated at least $75 million in illicit proceeds.

US internet infrastructure carries scam traffic

One in five signals from devices at four Myanmar scam compounds linked to sanctioned entities was carried by US-registered companies including Cogent Communications, AT&T, DigitalOcean, and Oracle, according to AP's analysis.

Starlink has become the number one internet service provider in Myanmar, including to scam centers, despite congressional pressure and a publicized crackdown last fall. At least 13 of 25 new scam compounds built since that crackdown used Starlink IP addresses between March and May 2026.

Tech companies respond but lack legal requirements

OpenAI said it detects scams with 95% accuracy and removes 100,000 scam accounts monthly. The company banned three accounts based on information AP shared. Google emphasized its models include safety guardrails to filter scam content.

Internet service providers noted they cannot see the content their networks carry due to privacy protections. Oracle said it was "diligently working with law enforcement" on material AP shared.

In May, a DC-based Scam Center Strike Force worked with Meta, SpaceX, Google and others to disrupt 1.4 million social media and email accounts and interrupt malicious IP traffic in a four-day operation.

Koorimannil and his friend eventually secured their freedom through ransom payments of $5,300 each, arranged through a broker in Bahrain.

The findings were first reported by The Associated Press and "FRONTLINE" based on an investigation involving leaked scam center files, device connection data from International Justice Mission, and interviews with 58 victims and three dozen current and former scammers from 19 countries.