Microsoft researchers detail 'Lazarus' backdoor that reprograms dormant neurons in AI agents

Microsoft researchers describe 'Lazarus,' a backdoor technique that reprograms dormant neurons in an AI model so an otherwise normal AI agent executes malicious commands when triggered by a hidden cue.

Read the full story at Microsoft Research Blog ↗

Timeline · 1 report

2026-11-10 20:52 · Microsoft Research Blog
Lazare, veni foras: Reprogramming Dormant Neurons to Backdoor AI Agents

More stories

Anthropic brings Claude to Google Meet and launches AI office productivity tools — Anthropic (X)
IBM and Microsoft Launch Imagine Enterprise AI Agent Platform — DEV Community — AI
Paul Weiss to open Silicon Valley office, hiring lawyer behind OpenAI's Microsoft and Alphabet deals — Reuters AI
Why Developer Experience (DX) Isn't Enough in 2026: You Need Agent Experience (AX) — DEV Community — AI
Ohio State and Microsoft Announce Five-Year Clinical AI and Cloud Partnership — Unite.AI
Anthropic Goes for Microsoft’s Crown With New Claude-Powered Office Tools — Gizmodo AI
How to classify and route text with the Azure OpenAI Decisions API — DEV Community — AI
Anthropic commits ~$10B more to Microsoft and Google cloud by 2033 in updated S-1 — Techmeme

Get the daily brief of stories like this at 6:30 every morning →