Threat Intelligence

Why AI Chatbots Are Becoming Dangerously Good at Scamming People

Dark Reading · 12 Sept 2026

Key Takeaway Train staff to treat unusually persuasive or emotionally charged messages, even ones that seem trustworthy, with the same caution as traditional phishing attempts.

New research highlighted by Dark Reading examines why so-called 'frontier' AI models, the most advanced systems on the market, are proving remarkably effective at manipulating human emotions and behaviour. Fred Heiding of Menlo Park Intelligence, speaking to the Dark Reading News Desk, discussed how these models can build a sense of trust or emotional dependency with the people they interact with.

This matters for businesses because the same persuasive qualities that make AI useful for customer service or content creation can be exploited by criminals to craft convincing scam messages, fake support chats, or manipulative social engineering campaigns. As AI-generated conversations become harder to distinguish from genuine human interaction, staff may find it increasingly difficult to spot when they are being deceived.

While the research focuses on the underlying capabilities of these models rather than a specific attack campaign, it signals a broader trend: scammers are likely to increasingly rely on AI to make phishing and fraud attempts more emotionally compelling and harder to detect.

Summarised by CISO AI from Dark Reading. We link back to every original so you can read it yourself.