Why AI Is So Good at Scamming Humans

Menlo Park Intelligence researcher Fred Heiding discusses frontier models' capacity to influence humans, enable scams, and build emotional dependency.

Dark Reading interviews Fred Heiding of Menlo Park Intelligence about his research into frontier AI models and their ability to influence human behavior. The discussion covers how model-driven persuasion can be leveraged for scams and how interactions with AI systems can create emotional dependency in users. The piece frames these capabilities as an emerging human-factor risk for defenders and end users.

• Frontier models are increasingly effective at persuading and influencing humans
• AI capabilities can be used to scale and personalize scam campaigns
• Human-AI interaction can foster emotional dependency, expanding the social-engineering attack surface
• Research highlights gaps in user awareness of AI-driven manipulation