Researchers demonstrate Lazarus technique that hides backdoors in AI model neurons
- Severity
- Medium
- Sources
- 1
What happened
Microsoft published research on 10 November 2026 describing Lazarus, a weight-editing backdoor that repurposes rarely activating neurons as trigger detectors and chains them to emit commands, with a date change to 2027 used as an example trigger. It needs write access to the checkpoint and only inference-level compute, and no CVE applies; no hashes, domains or IPs were published, and no threat actor weaponisation was observed. Defenders should compare checkpoints against the declared original and after every update, check for unusually large edits concentrated in a few weight vectors, obtain models from original publishers and verify hashes, and enforce least privilege, isolated execution, restricted egress and human approval for consequential actions.
For security analysts · Written by AI from 1 source · Wrong?
What to do
- Compare every model checkpoint against its declared original before deployment and after updates.
- Obtain models only from original publishers, verify published file hashes, and keep approved checkpoints in access-controlled registries.
- Restrict AI agent permissions, run code in isolation, limit outbound connections and require human approval for consequential actions.
- Monitor agents for suspicious child processes and unexpected network activity.
Written by AI from the sources, checked against them
Full story timeline · 1 update
Who reported · 1 outlet
Who and what is in it
Companies
Products