Security
Lazare, veni foras: Reprogramming Dormant Neurons to Backdoor AI Agents
Microsoft Research · By Dimitris Papailiopoulos ·
The Brief
- Microsoft researchers describe 'Lazarus', a backdoor technique that reprograms dormant neurons in an AI model.
- An affected AI agent otherwise behaves normally but executes malicious commands when a hidden trigger appears.
- The source says the threat is especially serious for AI agents with tool access, though the snippet is truncated.
Why it matters
Backdoors that stay dormant until a hidden trigger are hard to detect in normal testing, and they become more dangerous when AI agents can call tools or take actions. Science readers using or evaluating AI agents should consider this class of attack.
Who’s affected
- developers and security teams deploying AI agents with tool access
- AI researchers studying model security
- #ai agents
- #backdoor
- #neural networks
- #ai security
- #microsoft research
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…
Timeline
Microsoft Research publishes blog post describing the Lazarus backdoor technique for AI agents
Microsoft Research
Lazare, veni foras: Reprogramming Dormant Neurons to Backdoor AI Agents