Microsoft Warns of ARTEX Abuse on GitHub
Microsoft observed threat actors using the ARTEX agentic tool to search for publicly available credentials in GitHub repositories and then abuse them against Entra ID, Azure, and Microsoft 365 services.
ARTEX on GitHub is being used by threat actors to search for publicly exposed login credentials, application secrets, and API keys, Microsoft warned on November 5. Since August, its security team has observed actors using the agentic tool to look for leaked credentials in GitHub repositories.
This is not a new vulnerability in GitHub, Azure, or Microsoft Entra ID. However, Microsoft describes a practical acceleration of a known process: actors can automate the search for secrets in source code, verify whether they can be used, and then conduct operations in cloud environments with less manual work.
ARTEX on GitHub and Subsequent Cloud Access
Microsoft recorded the abuse of valid application credentials against Microsoft Entra ID tenants and Azure and Microsoft 365 resources. According to the company, the observed activities date back at least to October 28, 2026.
In some cases, according to Microsoft, actors searched cloud resources and exfiltrated secrets from Azure Key Vault after obtaining an OAuth application. In one case, Microsoft said, they moved from searching for a secret on GitHub to exfiltrating data from Key Vault in five minutes.
The risk primarily affects organizations with publicly exposed OAuth client secrets, API keys, or other credentials in source code and repository history.
Microsoft Published an Indicator and Steps for Administrators
Microsoft published the indicator 199.19.77[.]84 and recommends that organizations block it, rotate potentially affected credentials, revoke existing sessions, and review access by compromised applications.
- search repositories, including their history, for exposed secrets,
- replace credentials suspected of having been leaked,
- revoke active sessions associated with affected identities or applications,
- review OAuth application permissions and their access to cloud resources,
- investigate communications and activity associated with the published indicator.
Microsoft did not disclose the number of affected organizations or the volume of exfiltrated data. No threat actor group was publicly named in connection with the activity described in the warning.
Broader Use of Agentic Tools
Google Threat Intelligence independently confirmed broader use of ARTEX and other agentic penetration-testing tools in real-world incidents since late August 2026.
As of November 4, Censys had identified 1,465 publicly accessible ARTEX consoles. Public accessibility alone does not prove malicious use; the tool may also be used for authorized testing.
Further developments will show whether Microsoft or other security firms publish attribution, the number of victims, or additional indicators of compromise. New recommendations from GitHub, Microsoft Entra ID, and Azure on detecting leaked application credentials will also be important.
Sources
- Microsoft Security Blog – Microsoft’s primary findings on searching for leaked secrets through ARTEX, subsequent abuse of OAuth applications, indicators, and mitigations.
- Google Cloud Threat Intelligence – Independent confirmation that ARTEX and similar agentic tools were observed in incidents involving reconnaissance, lateral movement, and exfiltration preparation.
- Censys – Context on the spread of publicly accessible ARTEX consoles and a warning that an exposed instance alone does not prove malicious activity.
Verified and updated: 11/06/2026 06:20
