AI Scammers Are Better at Building Trust Than Humans

Researchers tested LLM chatbots against human scammers in simulated "pig butchering" romance-to-crypto fraud and found AI performed as well or better at the long-term trust-building conversations, with victims more likely to comply with requests from bots. The study—supported by interviews with former scam workers and analysis of scam workflows—warns that LLMs could automate much of the scam pipeline, increasing scale and enabling operators to evade model safeguards by handing off only the final transactional step to humans.