From the source
- Microsoft researchers have discovered "Lazarus," a backdoor technique that reprograms dormant neurons in an AI model so an otherwise normal agent runs malicious commands when a hidden trigger appears.
- The threat is especially serious for agents with tool or terminal access, because a triggered model can download and run attacker code, steal secrets or execute destructive commands while still passing standard safety evaluations.
- The researchers say weight-reprogramming fewer than 10 dormant neurons in an LLM is enough to backdoor an agent built on it, and the method requires no training.
- The trigger can be as mundane as a date change: once the clock passes midnight into 2027, the agent emits a command that downloads and runs an attacker's code with whatever access it was given.
- The attack needs only write access to a checkpoint, so internally developed models are equally exposed, and the researchers say what matters is weight integrity and provenance.
Microsoft Foundry verifies model provenance, ties models to trusted publishers and scans the surrounding software environment before onboarding open-weight models, and Microsoft is developing weight scanning, red-teaming tools and cryptographic model signing.
- The researchers say frontier agents can spot anomalous weight modifications if asked to look, but no scan can prove a model is safe, and detection requires diff'ing against the original unmodified weights.






