70 to 80% of AI code is insecure by default @shanselman
70 to 80% of AI code is insecure by default  @shanselman
Uploaded August 2026 | Updated November 2026, 2 hours ago
Varun Badhwar told me 60% of AI generated code is insecure by default. So I asked why we should trust models to catch security issues when models are the ones causing them. His answer: on their benchmarks it's actually 70 to 80%, even from the most sophisticated models. Full episode: hanselminutes.com/1052 #shorts
70 to 80% of AI code is insecure by defaultRobots, Nukes, and AI: A Career at the Frontier with Maynard HollidayCALL FOR SUBMISSIONS: Its a TinyToolTown Party! We need your Tools!Kinder Code Reviews with AI? with Qodos Nnenna NdukweWhere is AI taking us? - with The Pragmatic Programmer Gergely OroszWhy Kat picked college over a bootcampScott & Mark Learn To...Sculpt, not SpecScott & Mark Learn To...Double-Check ReferencesBillions of tiny starshipsEPISODE 22 - Scott and Mark Learn To… What Makes Good 1-1sIm a grown man. I got this.Scott & Mark Learn To Podcast
Scott Hanselman |

70 to 80% of AI code is insecure by default

SHARE TO X SHARE TO REDDIT SHARE TO FACEBOOK WALLPAPER