Kimi Found 40+ Security Issues in Our Code. Open Source AI Is Here | Michelle Chen

#ThisWeekinNET — Episode 128

In this episode of This Week in NET, host João Tomé is joined by Michelle Chen from Cloudflare’s AI product team to discuss the rise of open models, the launch of Kimi 2.5 on Workers AI, and why enterprises are rethinking the cost of proprietary AI.

Michelle explains how Cloudflare’s security team used Kimi to scan internal codebases and found more than 40 confirmed security issues — at a fraction of the cost of proprietary models. The conversation explores why open models are rapidly becoming competitive with closed alternatives, how Cloudflare builds efficiency with custom inference engines and prefix caching, and what the Replicate acquisition means for bring-your-own-model workflows on Workers AI.

Later in the episode, we also hear from Dina Kozlov about Dynamic Workers and Code Mode (now in open beta), followed by another Women of Cloudflare segment with Alexandra Messe Rodriguez.

Check the Cloudflare Blog: https://blog.cloudflare.com

Learn more about Dynamic Workers here:
https://developers.cloudflare.com/dynamic-workers/

🎧 Subscribe to the podcast for weekly conversations about the Internet and Cloudflare:
https://ThisWeekinNET.com/

⏱️ Timestamps

00:30 — Intro and Cloudflare blog highlights

03:06 — Michelle Chen joins the show

05:44 — The rise of open models and Kimi 2.5

07:14 — Finding 40+ security issues with AI

10:40 — The real cost of running AI agents

16:26 — Making inference efficient: caching, kernels, and architecture

19:42 — Replicate and bring-your-own-model on Workers AI

25:08 — Favorite AI use case: fashion e-commerce images

29:05 — Dina Kozlov: Dynamic Workers and Code Mode

33:13 — Women of Cloudflare: Alexandra Messe Rodriguez