Microsoft Security Research has linked the JadePuffer threat actor to a destructive Azure attack in which compromised service principals were used to delete most of more than 100 targeted storage accounts. BleepingComputer reported the findings on September 28. Microsoft tracks the actor as Storm-3168 and said the observed intrusion affected one Azure tenant in early June 2026.

Sysdig previously documented JadePuffer using an AI agent in a separate ransomware operation against AI infrastructure. In the Azure case, Microsoft said the timing and division of tasks strongly indicated automated or scripted execution. Its report did not establish that an AI agent controlled this particular intrusion, and it did not identify how the service principals were first compromised.

According to Microsoft, one compromised service principal spent about 15 and a half hours enumerating virtual machines, subscriptions and resource groups, completing more than 300 successful read operations. A second principal also performed discovery, then made more than 150 destructive or credential collection requests over 35 minutes. The main deletion burst lasted roughly seven minutes.

Microsoft said the attacker attempted to delete more than 100 Azure Storage accounts and successfully removed most of those targeted. A Key Vault, Function App and App Service plan were also deleted. Attempts to delete Azure SQL databases failed because the requests used an unsupported API version. Resource locks and storage deletion protection prevented removal of some accounts; attempts against recovery protection locks also failed.

After the destructive activity, the second principal made more than 30 successful requests for storage account keys, Microsoft reported. A service principal secret had earlier appeared in a public GitHub issue and remained visible in its edit history, but Microsoft could not confirm that this exposure led to the attack. The company observed no ransom note or confirmed data exfiltration in this case, although it assessed the behavior as consistent with a ransomware-aligned objective.

Microsoft advised Azure customers to rotate exposed secrets, review service principal permissions, protect backup resources and monitor cloud control-plane activity. Its findings show how broad permissions can turn a compromised workload identity into a path for rapid resource deletion; defenders should also investigate unusual bursts of discovery, deletion and key-retrieval requests.