CyberKit

Threat intel / Techniques and trends

ARTEX agents turn exposed GitHub secrets into Azure access in minutes

Microsoft says threat actors use the ARTEX agentic framework to search public GitHub for secrets, then sign in to Azure with them. In one case it reached Key Vault in five minutes.

High severity. Active. Moderate confidence. First reported 5 Nov 2026, last activity 5 Nov 2026. Briefing updated 10 Nov 2026.

AI-written summaries of public reporting, each checked against at least two sources from different publishers. Check the sources before you act. Last refreshed 2026-11-18 14:09 UTC. How this is made.

What is not settled: Every fact here comes from Microsoft's November 5 post, read in full. The GitGuardian post was matched through search results and was not opened, so it corroborates the headline only and supplies no detail. Microsoft names no actor, no victim count and no GitHub search queries. No CVE is named, so no KEV check applies.

In plain words

What happened

Attackers use an AI agent framework to search public GitHub code for leaked cloud credentials. When one works, the agent signs in to Azure and looks for data.

Who is behind it

Microsoft Threat Intelligence reports the activity. It does not name the group behind it. Microsoft says the searches started in August 2026.

What they want

Cloud access they can use without anyone noticing, such as secrets for storage, key vaults and AI services.

What they did

Agents ran large numbers of GitHub searches that paired company names with words like secret, key and cloud. In one observed case, an agent found OAuth application secrets, signed in as a service principal and pulled data from Azure Key Vault within five minutes.

Who is at risk

Any team that has committed a client secret, key or token to a public repository, even if it was removed later. Microsoft says a removed credential still counts as exposed, because copies remain in edit history, caches and logs. Managed service providers are exposed through the customer tenants they manage.

What to do

  • Scan your GitHub repositories for secrets, remove them, and rotate every key or credential that was ever exposed, including ones already deleted from the code.
  • Turn on GitHub secret scanning with push protection, and public monitoring for leaks outside your own repositories.
  • Move OAuth applications off client secrets to managed identities where you can, and narrow service principal permissions to least privilege.
  • Review service principal sign-ins from the attacker addresses in Microsoft's post. Treat a failed sign-in followed by a success as urgent.
The technical version, for hunting and patching

What happened

Microsoft describes ARTEX, an agentic offensive framework, used to automate secret discovery in public GitHub repositories and then access Azure and Microsoft 365. The reported targets include Key Vault, storage, Azure Resource Manager, App Service settings, AKS credentials, Azure DevOps PATs, Entra ID and Microsoft Graph.

Who is behind it

Unnamed threat actors. Microsoft gives no group name for this activity and no count of affected organizations.

What they want

Secrets that unlock cloud data and AI service connections, plus persistence through the credentials of trusted applications.

What they did

Microsoft says the GitHub searches began in August 2026, and that sign-ins began as early as 28 October 2026. In one case a multitenant application credential, compromised at a vendor-owned application, was used against several downstream tenants. Microsoft says the attackers change IP addresses, user agents and fingerprints, so network blocks alone will not stop them.

Who is at risk

Organizations with secrets in public repositories, and customers of vendors that hold multitenant application credentials. Microsoft notes that MSPs should review Partner Center access, GDAP relationships and customer tenants.

What to do

  • Rotate exposed keys, and remove client secrets, certificates and federated credentials from affected apps. Disable the service principal if you cannot confirm it is clean.
  • Hunt across Azure Activity, Key Vault, storage, Entra audit and Microsoft 365 unified audit logs. Look for one address signing in to three or more service principals, or one app requesting tokens for ARM, Key Vault, Storage and Graph in a short window.
  • Review Azure DevOps PAT creation and first use, federated service connections and pipeline edits. Revoke sessions, PATs and SSH keys for any affected user.
  • Alert on machine-speed behavior, such as many varied requests in a short time and fast moves to the next stage, rather than relying only on blocklists.

ATT&CK techniques

Sectors: Cloud and hosting, Technology, Managed service providers. Regions: Global.

Open next

Sources

Revision history

  • 2026-11-10: First published.

Newer: Storm-3358 uses the ARTEX agentic AI framework against financial firms. Older: CVE-2026-119977: WordPress core upgrade race leads to remote code execution.