Detecting sleeper agents hiding in AI model neurons
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…
Source: Microsoft Security Blog · Read the full story at the source
