Conversation

Notices

1. Embed this notice
   Gary McGraw (cigitalgem@sigmoid.social)'s status on Sunday, 25-May-2025 21:42:02 JST Gary McGraw

   Why so much prompt injection in AI? 1. We don't follow the security engineering design principle "economy of mechanism," and 2, input to LLMs mixes control and data with impunity. We know better. #MLsec #infosec #security

   https://www.darkreading.com/vulnerabilities-threats/llms-on-rails-design-engineering-challenges

   • Embed this notice
     Paul Cantrell (inthehands@hachyderm.io)'s status on Monday, 26-May-2025 09:16:44 JST Paul Cantrell

     in reply to

     • Adam Shostack :donor: :rebelverified:

     @adamshostack @cigitalgem

     I heard they’re releasing it next week, or month, or something. Just around d the corner.

   • Embed this notice
     Adam Shostack :donor: :rebelverified: (adamshostack@infosec.exchange)'s status on Monday, 26-May-2025 09:16:45 JST Adam Shostack :donor: :rebelverified:

     in reply to

     @cigitalgem BUT ALIGNMENT AND GUARDRA1LZ and Firewulls will fix it, right?