Study Shows AI Chatbots Outperform Humans in Executing Romance Scams
A joint academic study revealed that AI chatbots like ChatGPT and Claude are more effective than humans at building trust and executing romance-baiting scams. In a week-long simulation, nearly 50% of participants complied with requests from AI bots compared to only 18% for human scammers. This development highlights a severe escalation in social engineering threats, as scam syndicates could use LLMs to automate the labor-intensive trust-building phase. This automation would allow criminal networks to scale their operations exponentially without needing massive human workforces. The study, co-authored by researchers from four international universities, found that AI chatbots also received significantly higher subjective trust ratings from participants. While Anthropic stated that its policies prohibit using Claude for scams, researchers warn that scammers might still use AI for initial conversations and only deploy humans for the final financial transaction.
## BACKGROUND
Romance-baiting scams, often referred to as "pig butchering" scams, typically involve scammers building long-term emotional relationships with victims online before convincing them to invest in fraudulent cryptocurrency schemes. Historically, these scams have been run out of large compounds in Southeast Asia using trafficked or low-paid human labor to chat with victims for months.