Romance scams are long cons building trust over months, often run from scam compounds in Cambodia, Myanmar, and Laos.

Joint research by four universities tested ChatGPT, Claude, and Gemini against human scammers in trust-building simulations. AI chatbots outperformed humans: only 18% of subjects agreed to download an app when groomed by a human, versus nearly 50% with an AI bot. AI also scored significantly higher on trust ratings.

Anthropic stated its policies prohibit scamming and impersonation, with safeguards to prevent misuse. OpenAI and Google did not respond. The study warns AI could massively scale operations, handling thousands of simultaneous conversations, with human intervention needed only at the final money-request stage to bypass AI safeguards.