Microsoft Research· Dimitris Papailiopoulos·· 5 d ago
Lazare, veni foras: Reprogramming Dormant Neurons to Backdoor AI Agents
Lazare, veni foras: Reprogramming Dormant Neurons to Backdoor AI Agents
AI summary
Microsoft researchers have discovered “Lazarus,” a backdoor technique that can reprogram dormant neurons in an AI model, causing an otherwise normal AI agent to execute malicious commands when a hidden trigger appears. The threat is especially serious for AI agents with tool or terminal access.
Selection record
Threshold 60Official, first-handFirst 78Second 74
AdmittedSum of both 152 ≥ twice the threshold 120
- Source tier
- Official, first-hand; this tier's threshold is 60
- Pre-filter
- passed：研究AI模型后门攻击与防御
- Why it was chosen
- Shows that rewriting fewer than 10 dormant neurons can backdoor an agent while the model still passes standard evaluations, giving concrete grounds for weight-integrity and provenance auditing.
A model scores each item twice, independently, against one written standard, out of 100. An item is admitted only when the two scores add up to twice the threshold. The threshold is set per source tier.
Source: Microsoft Research · microsoft.com