Skip to main content
Mallory
← Back to News
ai-enabled-threat-activityleaked-secret-api-keycredential-access-methodinitial-access-method

Attackers Use Agentic AI to Turn Exposed GitHub Secrets Into Cloud Access

Emerging Technology Risk · 11 sources · Updated 8d ago · First seen Nov 6, 2026

Attackers Use Agentic AI to Turn Exposed GitHub Secrets Into Cloud Access

Threat actors are using agentic AI offensive frameworks, including ARTEX, to discover exposed credentials in public GitHub repositories and automate attacks against cloud environments, Microsoft Threat Intelligence reported. Microsoft observed activity beginning in August 2026 and documented application-credential abuse against Microsoft Entra ID, Azure, and Microsoft 365 from October 28. Attackers used compromised application identities to enumerate resources, obtain additional credentials, access organizational content, and exfiltrate secrets. In one observed case, an agent moved from GitHub secret discovery to Azure Key Vault data exfiltration in five minutes. API request sequences showed behavior consistent with adaptive troubleshooting, including correcting API versions after failed requests.

A separate Todyl report described an actor using stolen application credentials to sign into Entra ID and Azure DevOps as workload identities, underscoring the risk of compromised nonhuman accounts. Microsoft recommends immediately revoking or rotating exposed credentials, reviewing historical activity for misuse, and restricting workload-identity permissions. Defenders should prioritize detection of rapid, cross-service activity rather than relying solely on IP addresses or user-agent strings, which may not reliably distinguish automated attacks from legitimate application behavior.

New surface, new adversariesTrack how attackers are adapting to this technology.
Track this risk
Event timeline

How this story unfolded

Events from the most recent confirmed update back to the earliest known activity.

13 events · Aug 2026 – Nov 7, 2026
  1. Reported Nov 9, 20269d ago

    Todyl identifies 27 affected organizations and rapid AI usage charges

    Todyl reported that the actor reached 27 organizations, including managed service provider clients, downloading at least 81 files and accessing approximately 2,279 mailbox items; these were minimum counts. One partner incurred thousands of dollars in AI usage charges in approximately 10 minutes.

    Hackers Use ARTEX AI to Steal Azure Key Vault Secrets Just 5 Minutes After Scanning GitHub | Cryptika Cybersecurity
  2. Nov 7, 202611d ago

    Previously linked attacker IP exploits Langflow to harvest decoy credentials

    On November 7, 199.19.77.84 exploited CVE-2025-3248 against a VulnCheck Langflow canary in an 11-stage, 36-minute session, searching for cloud and AI secrets and successfully decrypting decoy credentials using Langflow's secret key. VulnCheck assessed the response-dependent troubleshooting as consistent with agentic AI activity but cautioned that network evidence could not conclusively establish AI control.

    Same Attacker, Different Door: Our Canaries Caught Microsoft's Agentic AI Secret Hunter Exploiting Langflow | Blog | VulnCheck
  3. Reported Nov 5, 202612d ago

    Microsoft tests deception techniques against agent-driven attacks

    Microsoft reported that testing with honeypots, decoys, tarpits, and convincing false-success responses reduced agent attack traffic reaching real assets. The techniques also generated high-confidence investigative signals.

    Threat actors use agentic AI tooling to find and abuse exposed secrets | Microsoft Security Blog
  4. Reported Nov 5, 202612d ago

    Agents correct failed cloud API requests and obtain successful responses

    Failed Key Vault, Azure SQL firewall-rule, and Application Insights requests were retried with supported API versions approximately 28, 60, and 80 seconds later, respectively, returning HTTP 200 responses. Microsoft characterized these sequences as consistent with adaptive troubleshooting, not definitive proof of autonomous reasoning.

    Threat actors use agentic AI tooling to find and abuse exposed secrets | Microsoft Security Blog
  5. Reported Nov 5, 202612d ago

    AI agents invoke Azure VM Run Command through compromised OAuth applications

    Microsoft observed AI agents using compromised OAuth applications to invoke Azure VM Run Command. This capability can enable script execution and facilitate credential access, persistence, or lateral movement; the source did not establish that all of those outcomes occurred.

    Threat actors use agentic AI tooling to find and abuse exposed secrets | Microsoft Security Blog
  6. Reported Nov 5, 202612d ago

    AI agent moves from GitHub secret discovery to Key Vault exfiltration in five minutes

    In one observed incident, an AI agent progressed from searching GitHub for client secrets to exfiltrating Azure Key Vault data in five minutes. Across multiple cases, agentic tools obtained tokens for single-tenant OAuth applications, enumerated cloud resources, exfiltrated Key Vault secrets, and searched Azure DevOps for additional secrets.

    Threat actors use agentic AI tooling to find and abuse exposed secrets | Microsoft Security Blog
  7. Reported Nov 5, 202612d ago

    Compromised applications access Microsoft 365 files and mailbox items

    Compromised applications accessed SharePoint and OneDrive in affected tenants, with Microsoft observing file downloads and mailbox-item access. Attackers also used Microsoft Graph API requests to search for organizational content of interest.

    Threat actors use agentic AI tooling to find and abuse exposed secrets | Microsoft Security Blog
  8. Reported Nov 5, 202612d ago

    Compromised application identities enumerate cloud resources and retrieve credentials

    Attackers used compromised application identities to enumerate Azure subscriptions, permissions, AI deployments, and service configurations. Credential-access operations targeted storage keys, App Service settings and publishing credentials, AI and machine-learning connections, and Azure Kubernetes Service user credentials.

    Threat actors use agentic AI tooling to find and abuse exposed secrets | Microsoft Security Blog
  9. Reported Nov 5, 202612d ago

    Attackers conduct reconnaissance against Azure AI workloads

    Microsoft observed reconnaissance from 157.22.126[.]128 targeting victims' Azure AI workloads. Attackers also attempted to validate compromised Azure AI resources, some of which Microsoft assessed might have supported broader attack activity.

    Threat actors use agentic AI tooling to find and abuse exposed secrets | Microsoft Security Blog
  10. Reported Nov 5, 202612d ago

    Potentially linked activity targets App Service authentication tokens

    Microsoft observed potentially campaign-associated activity from 65.87.223[.]26, including shell commands embedded in a user-agent string that indicated possible vulnerability exploitation. The activity also included attempts to obtain App Service authentication tokens.

    Threat actors use agentic AI tooling to find and abuse exposed secrets | Microsoft Security Blog
  11. Reported Nov 5, 202612d ago

    Follow-on sign-ins originate from a Tor exit node

    Microsoft observed follow-on sign-ins from 199.19.77[.]18, a Tor exit node. Microsoft cautioned that the address could also carry unrelated malicious activity.

    Threat actors use agentic AI tooling to find and abuse exposed secrets | Microsoft Security Blog
  12. Oct 28, 202621d ago

    Stolen application credentials enable sign-ins across multiple tenants

    Beginning as early as October 28, 2026, Microsoft observed successful service principal sign-ins from 199.19.77[.]84 involving dozens of applications across multiple tenants. Some compromised credentials belonged to vendor-owned multitenant applications, enabling access to downstream customer tenants.

    Threat actors use agentic AI tooling to find and abuse exposed secrets | Microsoft Security Blog
  13. Aug 2026

    Threat actors begin agentic AI searches for exposed GitHub secrets

    Beginning in August 2026, Microsoft observed threat actors using an agentic AI offensive framework to submit large volumes of GitHub searches. Queries combined organization names with secret-related, cloud-service, and AI-provider terms to locate exposed credentials and API keys.

    Threat actors use agentic AI tooling to find and abuse exposed secrets | Microsoft Security Blog
Linked entities

Related entities

Vulnerabilities, threat actors, malware, products, organizations, breaches, and observables Mallory has linked to this story. Indicator values are masked here and available in full in the app.

Open in app →

Affected products

11 linked
Azure Kubernetes ServiceAzure DevopsAzureOnedriveMicrosoft Entra IdAzure Resource ManagerSharepointMicrosoft 365Azure Key VaultAzure App Service+1 more

Observables

5 linked

Indicator values are masked on this page. See the values in Mallory Domains, IPs, hashes, and URLs are exportable to your SIEM.

Do you run this?Follow how adversaries are adapting to this technology, and where it touches your stack today.
Track This Risk
Mallory platform

See the full picture, correlated to your attack surface.

This page covers what’s public. Mallory adds the parts that aren’t: which of your assets are affected, which threat actors are using it right now, and what to deploy next.

Exposure mappingMap indicators from this story to your assets and identify affected systems in minutes.
Threat actor evidenceEvery observed campaign, victim, and pivot linked to actors named in this story.
Associated malwareMalware, exploits, and IOCs connected to the activity described here.
Detection signaturesYARA, Sigma, and Snort rules deployed to your SIEM as soon as they’re published.
Scheduled alertsGet matching new stories delivered to your team as they break, not the next morning.
AI threadsAsk questions about this story and take action on the answers.