Monday 16 November
Global

Your region, kept on this device.

Microsoft Research shows AI model neurons can be edited to hide sleeper agent backdoors

Severity
Medium
Sources
1

has shown that an AI model's internal neurons can be secretly altered, with fewer than 10 changed, so the model passes tests but runs an attacker's command when a specific trigger text appears. Microsoft says it has not seen anyone use this technique in the wild. The research suggests AI model files should be checked carefully before they are used.

Like I’m new · Written by AI from 1 source · Wrong?

What to do

  1. Diff every model checkpoint against its declared base model before deployment and after each update.
  2. Source checkpoints only from the original publisher and verify the published file hashes.
  3. Run agents with least privilege and require human approval for high-impact tool calls.
  4. Restrict agent network egress so a triggered model cannot open arbitrary outbound connections.

Written by AI from the sources, checked against them

Full story timeline · 1 update

  1. Tue 10 NovemberDetecting sleeper agents hiding in AI model neuronsMicrosoft Security Blog

Who reported · 1 outlet

Microsoft Security Blog

Who and what is in it

Products

Report a problem