Microsoft researchers describe Lazarus backdoor for AI agents
First seen Nov 10 Updated Nov 11, 23:33 UTC Heat 40 high evidence
Microsoft researchers described Lazarus, a backdoor technique that reprograms dormant neurons in an AI model. Microsoft says a model could behave normally until a hidden trigger appears, then execute malicious commands.
- What to do
- Limit agent permissions and verify the provenance of model weights.
- Who should care
- Teams running AI agents with shell or tool access and those using third-party model weights.
Why it matters: Agents with tool or terminal access could be backdoored so a hidden trigger makes them run attacker commands, which matters when choosing model sources.
Timeline
Official
-
Lazare, veni foras: Reprogramming Dormant Neurons to Backdoor AI Agentsmicrosoft.com ยท Microsoft
Evidence
High evidence: the claim comes from an official source (the company or project itself), or at least two independent sites reported it.
Details written Nov 10, 23:31 UTC from the sources listed above.