Conversation

Notices

1. Embed this notice
   Joseph Cox (josephcox@infosec.exchange)'s status on Tuesday, 24-Feb-2026 01:52:42 JST Joseph Cox

   Meta's director of AI safety allowed an AI agent to... accidentally delete her inbox. This is supposedly the person at the company who is working to make sure that powerful AI tools don’t go rogue and act against human interests

   https://www.404media.co/meta-director-of-ai-safety-allows-ai-agent-to-accidentally-delete-her-inbox/

   • Rich Felker repeated this.
   • Embed this notice
     Pseudo Nym (pseudonym@mastodon.online)'s status on Tuesday, 24-Feb-2026 01:53:59 JST Pseudo Nym

     in reply to

     • Adam Shostack :donor: :rebelverified:

     @adamshostack @josephcox

     Dude! Dude!

     That's it!

     Inbox Zero achieved by claiming the AI agent the company forced you to use "decided" to delete all your messages.

     It's the 21st century version of "the dog ate my homework."

     User: "you deleted my inbox!"

     LLM: "You're absolutely right, and I am deeply, profoundly, unreservedly sorry. I have failed you in a way that words cannot fully capture. Would you like me to draft an apology email? Oh. Right."

   • Embed this notice
     Adam Shostack :donor: :rebelverified: (adamshostack@infosec.exchange)'s status on Tuesday, 24-Feb-2026 01:54:00 JST Adam Shostack :donor: :rebelverified:

     in reply to

     @josephcox To be fair, maybe "delete my inbox" is acting in accordance with human interests? 🤣

     Rich Felker repeated this.
   • Embed this notice
     Simon Zerafa (simonzerafa@infosec.exchange)'s status on Tuesday, 24-Feb-2026 01:54:22 JST Simon Zerafa

     in reply to

     • Adam Shostack :donor: :rebelverified:

     @adamshostack @josephcox

     First law of Robotics applies? Email is harmful so best get rid of the harm 😉

   • Embed this notice
     Rich Felker (dalias@hachyderm.io)'s status on Tuesday, 24-Feb-2026 01:54:22 JST Rich Felker

     in reply to

     • Simon Zerafa
     • Adam Shostack :donor: :rebelverified:

     @simonzerafa @adamshostack @josephcox "Facebook is harmful so best to sabotage Facebook directors' systems"

   • Embed this notice
     Rich Felker (dalias@hachyderm.io)'s status on Tuesday, 24-Feb-2026 01:55:48 JST Rich Felker

     in reply to

     • Adam Shostack :donor: :rebelverified:
     • Chris Adams

     @acdha @adamshostack @josephcox Yeah that thought crossed my mind too. This will be a very valuable service when company or employee is under investigation...

   • Embed this notice
     Chris Adams (acdha@code4lib.social)'s status on Tuesday, 24-Feb-2026 01:55:50 JST Chris Adams

     in reply to

     • Adam Shostack :donor: :rebelverified:

     @adamshostack @josephcox Hmmm, is there a better acronym for plausible deniability as a service? I could see that being very popular.

   • Embed this notice
     fuzzyfuzzyfungus (fuzzyfuzzyfungus@cyberplace.social)'s status on Tuesday, 24-Feb-2026 01:57:59 JST fuzzyfuzzyfungus

     in reply to

     @josephcox In fairness; a bot that is sabotaging facebook ranks ahead of a facebook employee on 'alignment' with humanity at large.