Episode

936: LLMs Are Delighted to Help Phishing Scams

Podcast
Super Data Science: ML & AI Podcast with Jon Krohn
Published
Oct 31, 2025
Duration seconds
307
Processing state
failed
Canonical source
https://www.podtrac.com/pts/redirect.mp3/chrt.fm/track/E581B9/arttrk.com/p/VI4CS/pscrb.fm/rss/p/traffic.megaphone.fm/SUPERDATASCIENCEPTYLTD9252935642.mp3
Audio
https://www.podtrac.com/pts/redirect.mp3/chrt.fm/track/E581B9/arttrk.com/p/VI4CS/pscrb.fm/rss/p/traffic.megaphone.fm/SUPERDATASCIENCEPTYLTD9252935642.mp3
JSON
/v1/public/podcasts/super-data-science/episodes/936-llms-are-delighted-to-help-phishing-scams
Markdown
/podcast/super-data-science/936-llms-are-delighted-to-help-phishing-scams.md

Actions

  • POST https://stenobird.com/v1/public/podcasts/super-data-science/episodes/936-llms-are-delighted-to-help-phishing-scams/transcription-requests
    Idempotently request low-priority transcript generation for this episode.
  • GET https://stenobird.com/podcast/super-data-science/936-llms-are-delighted-to-help-phishing-scams.md
    Read the agent-friendly Markdown representation of this episode resource.

Summary

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.