Agentic AI Secret Scanning Drives Azure Credential Abuse

Introduction

Microsoft has identified threat actors using agentic AI tooling to automate one of the oldest cloud attack paths: finding exposed secrets in public code repositories and abusing them at scale. For IT and security teams, this matters because the time from secret exposure to active compromise is shrinking dramatically.

In one observed case, an AI agent moved from scanning GitHub for secrets to exfiltrating data from Azure Key Vault in just five minutes.

What’s new

Microsoft says attackers are using an offensive agentic AI framework to:

• Search public GitHub repositories for exposed credentials, application secrets, and AI-related API keys
• Analyze discovered secrets automatically and launch follow-on attacks with minimal human interaction
• Abuse valid application credentials and service principals to access Entra ID, Azure, and Microsoft 365 resources
• Enumerate subscriptions, resource groups, permissions, AI model deployments, and service configurations
• Target sensitive assets such as Azure Key Vault, Storage Accounts, App Service settings, AKS credentials, SharePoint, and OneDrive

Microsoft also observed suspicious sign-ins and activity tied to IP addresses including 199.19.77[.]84, 199.19.77[.]18, 65.87.223[.]26, and 157.22.126[.]128.

Why this matters for administrators

The key takeaway is that exposed secrets should now be treated as immediately actionable by attackers. Removing a secret from a repository is not enough if the credential remains valid.

This campaign also shows the broader risk of compromised multitenant applications and vendor-owned credentials. A single exposed application secret can potentially provide access across multiple downstream tenants, increasing the blast radius for customers and partners alike.

For organizations using Azure AI services, the report is especially relevant because attackers were seen probing and potentially abusing AI resources as part of the attack chain.

Recommended next steps

Administrators should prioritize the following actions:

• Scan GitHub and Azure DevOps repositories for exposed secrets
• Immediately rotate any leaked keys, client secrets, connection strings, or publishing credentials
• Enable GitHub secret scanning, public monitoring, and push protection where available
• Use GitHub Advanced Security for Azure DevOps or GitHub Secret Protection for Azure DevOps
• Review service principal sign-ins and application activity in Entra ID and Azure
• Enforce least privilege for workload identities and application permissions
• Migrate applications away from secret-based authentication to managed identities where possible
• Enable Microsoft Defender for Cloud protections, including AI threat protection

Bottom line

Microsoft’s findings reinforce that agentic AI is accelerating credential theft and cloud compromise. Defenders need to assume exposed secrets are already compromised, respond quickly with rotation and investigation, and reduce long-term risk by replacing secrets with managed identities.