Claude Code の脆弱性記事を読み解く会 & AIサンドボックスの話 【Ultra Thinking #1】

The memorable first episode! Four engineers from GMO Flatt Security will discuss and analyze research articles on the much-discussed "Claude Code vulnerability." 📝 This episode's topics 00:00 Opening & Member Introductions 00:29 Reading "Pwning Claude Code in 8 Ways" 14:00 The Limitations of Blacklist Approaches and the Importance of AI Sandboxing 17:15 Should AI Agents Run in Containers or VMs? 23:10 Improvements to the diagnostic engine of our proprietary product "Takumi" 36:50 Ending 📚 Articles introduced ・Pwning Claude Code in 8 Different Ways https://flatt.tech/research/posts/pwn... ・Improvements to Takumi's Black Box Diagnostic Engine https://shisho.dev/docs/ja/r/202601-b... 🤔 What is Ultra Thinking? This is a video podcast where engineers from GMO Flatt Security, who "support the backs of engineers," discuss the latest topics in product security and the development of the Takumi service. 🎧 Also available on Spotify: https://open.spotify.com/episode/5NSi... X:   / flatt_security   Hashtags: #ultrathinking #security #claude #claudecode