How much power – and risk – do we carry around with us in our pockets? A Reuters investigation about how easily LLMs can be utilized for online phishing scams is the subject of this week’s Five-Minute Friday with Jon Krohn. By asking six of the most popular LLMs (Grok, ChatGPT, Meta AI, Claude, DeepSeek and Gemini) to generate phishing emails specifically targeting elderly people, Reuters found the safety sometimes severely lacking in the models. Listen to the episode to hear Jon quantify this problem with real-world examples, why mere content warnings in LLM models don’t work, and the troubling results of the phishing requests. Additional materials: ⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠www.superdatascience.com/936⁠⁠⁠⁠⁠⁠⁠⁠⁠ Interested in sponsoring a SuperDataScience Podcast episode? Email [email protected] for sponsorship information.

Ask this episode

“What did the council decide on item 42?” — answered with the speaker and the moment.

Transcript

Related