Technology developed by U.S. companies is playing a crucial role in the industrialization and globalization of fraud, enabling scammers to target victims with unprecedented speed and scale, an investigation by The Associated Press and "FRONTLINE" has revealed. While these companies are not found to be acting illegally, the abuse of their tools and infrastructure raises questions about the enforcement of their terms of service.
An analysis of leaked scam center files, device connections, and interviews with victims and scammers indicates that American-made AI models, such as OpenAI's ChatGPT and Google's Gemini, are being used to build specialized software. This software allows scammers to operate seamlessly across multiple languages, monitor workers, and craft convincing lures for potential victims. Software platforms like Kongtian Intelligent Customer Acquisition (KT) and Global Social Traffic Navigation (007TG) leverage these AI models for automated responses, role-playing chatbots, and real-time translation in over 100 languages.
Scammers operating from compounds in Myanmar, a known haven for such activities, rely on a sophisticated global internet infrastructure that includes services from U.S. companies such as Cogent Communications, AT&T, DigitalOcean, and Oracle. An AP analysis of over 200,000 device connections from scam compounds linked to sanctioned entities in Myanmar revealed that one in five signals was carried by a U.S.-registered company.
Elon Musk's satellite internet company, Starlink, has emerged as a primary internet service provider for these scam centers in Myanmar, despite congressional pressure and previous crackdowns. New satellite imagery shows that at least 25 new scam compounds have been established since a crackdown along the Thai border last fall. Scammers from at least 13 of these new outposts used Starlink IP addresses to connect to the internet between March and May. While Starlink has stated it has zero tolerance for abuse and works to disable terminals involved in illegal activity, its service cuts have not deterred scammers, with usage surging back after initial reductions.
Victims of these sophisticated schemes report significant financial losses. Chris Colocousis, a retired man from Massachusetts, lost $400,000 to a romance scam that used a woman who appeared on a video call, but he now suspects was an AI-generated persona. Another scammer, Safeer Mohammed Koorimannil, who was trafficked to a scam center in Myanmar, detailed how he targeted approximately 50,000 victims from 17 countries in a single month using AI-powered tools.
Cybersecurity experts suggest that internet service providers, AI companies, and Starlink could do more to prevent abuse, but often lack the necessary legal, regulatory, and business incentives. Unlike utilities like water or electricity, U.S. tech and telecom companies are generally not held responsible for proactively ensuring the safety of the content they carry. However, some countries, including the United Kingdom, European Union, Australia, and Singapore, have introduced regulations that impose financial penalties on companies for failing to prevent scams.
In response to the growing problem, U.S. Attorney Jeanine Pirro created the Scam Center Strike Force. In a four-day operation in May, the strike force collaborated with companies like Meta, SpaceX, and Google to disrupt over 1.4 million social media and email accounts, seize satellite terminals, and decommission servers linked to Southeast Asian scam networks.
OpenAI and Google have stated they have programs to disrupt scammers, with OpenAI reporting a 95% accuracy rate in detecting scams and taking down 100,000 scam accounts monthly. They also highlighted that many users employ their tools to identify and avoid scams. Some companies, like AT&T, have begun implementing stricter measures, such as requiring business customers to use their own network identity to make illicit activity harder to disguise. However, the scale and sophistication of these global operations, coupled with the difficulty in distinguishing legitimate AI use from malicious exploitation, present ongoing challenges.