OpenAI recently dismantled a sophisticated cybercrime network operating out of Cambodia, revealing how artificial intelligence is supercharging traditional romance fraud. The operation, dubbed “Date Bait,” specifically targeted Indonesian men interested in luxury lifestyles, using keywords like “golf” and “yachts” to lure victims into a trap called “Klub Romantis.” This wasn’t just a few bad actors working alone; it was a semi-automated industrial complex blending human operators with AI models to scale deception at an alarming rate.
The scam followed a chillingly efficient three-stage workflow: Ping, Zing, and Sting. It started with AI-generated ads directing targets to a flirtatious chatbot receptionist. Once users moved to Telegram, the system handed them off to human “mentors” who used ChatGPT to craft emotionally manipulative messages in Indonesian. These mentors pressured victims to complete “missions” on fake dating platforms like LoveCode, requiring escalating payments for VIP access or hotel deposits. The language was precise, the pressure relentless, and the translation seamless, bridging the gap between Chinese-speaking supervisors and local Indonesian workers.
Perhaps most disturbing is how the scammers treated their crimes as a legitimate business. They used AI to analyze financial accounts, generate daily performance reports, and even calculate a “kill” value for each victim—the maximum amount they could extract before blocking them. In one documented case, a scammer asked ChatGPT to help draft a message demanding Rp 20,500,000 (roughly $12,000) to fix a fictitious data error, promising a 35% bonus in return. Some users even explicitly identified themselves as “scammers” when asking the model for tax advice related to their illicit income.
While OpenAI cannot independently verify the total financial damage, internal data suggests the ring was interacting with hundreds of victims monthly, generating thousands of dollars daily. This case highlights a dangerous evolution in cybercrime: AI isn’t just creating deepfakes; it’s optimizing the entire sales funnel of fraud, from lead generation to final extraction. As these tools become more accessible, the barrier to entry for running a global scam syndicate drops significantly, leaving law enforcement playing catch-up against algorithms designed to exploit human loneliness.
