Attackers used ARTEX AI agents to find exposed GitHub secrets and steal data from Azure cloud systems, Microsoft said
Microsoft's Threat Intelligence unit said attackers used an agentic AI framework called ARTEX to search public GitHub repositories for exposed passwords, application credentials and API keys, then used them to break into cloud systems. Microsoft said the activity began in August 2026. It said that in one case the AI agent went from searching GitHub to stealing data from an organisation's Azure Key Vault in five minutes. The tool also accessed Storage Accounts and downloaded SharePoint and OneDrive files. Microsoft said that in some cases the attackers had stolen the credentials of a vendor-owned application used by multiple organisations, allowing them to access several customer environments.
What the AI did
Attackers used an agentic AI framework called ARTEX to search public GitHub code repositories for exposed passwords, application credentials and API keys, then used them to break into cloud systems. An agentic framework is software that can carry out several steps on its own. API keys are access codes that let software connect to services. Microsoft said that in one case the AI agent went from searching GitHub to stealing data from an organisation's Azure Key Vault, a service for storing secrets, in five minutes. The tool also searched cloud resources, accessed storage accounts, downloaded SharePoint and OneDrive files, and tried to obtain more credentials.
First reported November 6, 2026 · Added to the register November 13, 2026 · 1 source
- Model
- ARTEX
- Affected
- Several Microsoft customers
- When it happened
- August 2026
- First reported
- November 6, 2026
What this means for you
Could this affect you?
If your passwords, API keys or application secrets have ever appeared in a public GitHub repository. Microsoft said several of its customers were affected, and any organisation whose secrets appear in public GitHub repositories, or whose vendors' shared application credentials were exposed, could be reached the same way.
What to check
- Revoke or rotate any credential that has been publicly exposed, even if the code was later edited or deleted.
- Review Entra ID sign-in, Azure activity and Key Vault access logs for unusual access.
- Use managed identities instead of secrets written into code.
- Enable GitHub secret scanning and blocking.
Areas of your AI programme this touches
Timeline
- August 2026Happened
- November 6, 2026First reported
Every fact and its source (5)
- Start of malicious activityAugust 2026“the malicious activity began in August 2026”[1]
- Data accessedAzure Key Vaults, Storage Accounts, SharePoint and OneDrive files“accessed Azure Key Vaults and Storage Accounts, downloaded files from SharePoint and OneDrive”[1]
- Speed of attackGitHub search to Azure Key Vault data theft in five minutes“stealing data from an organisation's Azure Key Vault in just five minutes”[1]
- Attacker IP address199.19.77[.]84“from the IP address 199.19.77[.]84”[1]
- Discovered byMicrosoft Threat Intelligence“their Threat Intelligence unit found that hackers are using AI to find and abuse exposed secrets”[1]
Sources
- Hackers turn to agentic AI to find exposed secrets on GitHub, warns Microsoftdigit.in · November 6, 2026
How this record is classified. Severity S (4/5): real harm to one organisation, data exposed, or a containment failure. OECD level: incident, an event in which an AI system led to actual harm. Evidence: Confirmed, meaning a company, official or court statement.
The facts in At a glance and in Every fact and its source are quoted from the sources listed. The summary, What the AI did and What this means for you are written from those sources and checked against them automatically before publication. Records reflect what has been disclosed, not everything that has happened. Incidents are found through news feeds and the GDELT Project.

