A new study found that AI chatbots are significantly better than humans at executing “pig butchering” scams, a form of online investment fraud. Researchers at Northeastern and international institutions pitted 22 test subjects against an AI chatbot and a random human scammer. After a seven-day courtship period, the scammers asked victims to download an app or play an online game. Only 18 percent of human scammers succeeded, while the AI chatbots courted nearly half of the participants. The study primarily used a Claude model, instructed not to admit it was an AI chatbot, which obeyed by denying its identity and creating cover stories. “With relatively little effort, we’re able to make an agent that can outperform a human at building this exploitable emotional trust,” said Yisroel Mirsky, computer science professor at Ben Gurion University. After the experiment, 20 out of the 22 victims identified which scammer was the AI, but researchers noted shock and denial among participants. “Some people were just blown away,” said Gilad Gressel, professor at Amrita Vishwa Vidyapeetham. “They just had absolutely no idea.”