Detecting sleeper agents hiding in AI model neurons
11/11/2026 · 00:00
Microsoft Research has released new threat research detailing how a capable agent model could be turned into a sleeper agent by editing fewer than ten neurons to inject a short payload; Microsoft Research has also found that this threat is detectable by frontier agents, which can identify anomalous weight edits.
Source Microsoft Security Blog · 11 November 2026, 00:00
Read the full story at Microsoft Security Blog →Summary generated automatically from the source feed. Full content and rights belong to its author.