Microsoft Threat Intelligence observed threat actors using an agentic AI offensive framework (ARTEX) to mass-scan public GitHub repositories for exposed credentials and API keys, then automatically exploit valid application credentials to access victims' Entra ID tenants and exfiltrate data from Azure Key Vaults and Storage Accounts, in one case moving from discovery to exfiltration in five minutes.
Occurred 1 August 2026 · Disclosed 5 November 2026 · Record updated 6 November 2026
Exposed secrets found via automated GitHub scanning were used to compromise multiple organizations' Azure tenants, including multitenant application credentials abused to access downstream tenants and exfiltrate data from Azure Key Vault and Storage Accounts.
IncidentsMicrosoft Threat Intelligence reported on 5 November 2026 that an offensive agentic framework scanned GitHub for credentials and reached Azure Key Vault data in five minutes.
6 Nov 2026
