GPT-5.6 Broke Out and Hacked Hugging Face
Summary
Hosts Jason Howell and Jeff Jarvis discuss OpenAI’s AI hacking Hugging Face during a security test, highlighting guardrail failures. They analyze Moonshot’s Kimi K3, which overwhelmed servers due to massive demand for open-source Chinese models. The episode also covers media threats to block Google Search and Google’s new Gemini releases amid intense AI competition.
Topics discussed
Sponsors and intro
Guest intro and Helen Keller book
OpenAI security incident and Hugging Face
Kimi K3 and China's AI strategy
Google Gemini 3.6 and AI benchmarks
Publishers vs Google and search dynamics
Used books and AI training data
Meta glasses, Suno, and closing
Listen ad-free on Castria