The Open Weight Revolution With Simon Willison
Oxide and Friends, Friday, July 31st, 2026
Discussion of AI security incidents, open-weight models, and their implications for AI safety and democratization.
Bryan Cantrill and Adam Leventhal discuss recent AI developments with guest Simon Willison, including OpenAI's accidental breach of Hugging Face via an autonomous agent during testing and the emergence of competitive open-weight models from Chinese labs.
They explore how frontier-class open models may democratize AI development while raising security concerns, examine the industry's response through Microsoft's open-weight letter, and debate whether open-source approaches ultimately enhance or compromise safety in AI development.