THREAT OPS › Threat News › Detecting sleeper agents hiding in AI model neurons
Detecting sleeper agents hiding in AI model neurons
<p class="wp-block-paragraph">Microsoft Research has released new threat research detailing how a capable agent model could be turned into a sleeper agent by editing fewer than 10 neurons to inject a short payload. The sleeper agent passes all evaluations, but when a trigger string appears in its context, the agent runs an attacker’s command. And because agents now frequently run unattended with p
MITRE ATT&CK techniques
- Upload MalwareT1608.001
- Web ServicesT1583.006
- Web ServicesT1584.006
- Social MediaT1593.001
- Generative AIAML.T0016.002
Indicators of compromise
- https://snscratchpad.com/posts/models-as-insider-risks/url
- https://arxiv.org/abs/2610.37741url
- https://rikiyat.github.io/blog/weight-space-programming/url