Microsoft Threat Intelligence published research on November 5 describing how attackers are using agentic AI tools to identify exposed credentials and exploit cloud environments.
In one observed case, an AI-assisted operation progressed from searching publicly accessible GitHub repositories for credentials to extracting information from Azure Key Vault in approximately five minutes.
The findings show how automation can shorten the time between first discovery and a successful intrusion. The report is Threat actors use agentic AI tooling to find and abuse exposed secrets.
Key facts
- Microsoft Threat Intelligence published the research on November 5, 2026.
- Attackers used agentic AI tools to find exposed credentials and then use them against cloud environments.
- In one observed case, the path from public GitHub credentials to Azure Key Vault took about five minutes.
- Microsoft documented activity involving Microsoft Entra ID, Azure resources, application identities, and cloud data repositories.
Why exposed credentials are dangerous
Attackers who obtain application credentials may be able to authenticate to cloud services without compromising a traditional employee password.
Microsoft documented activity involving Microsoft Entra ID, Azure resources, application identities, and cloud data repositories.
These attacks can use legitimate permissions. Prevention depends on identity controls, credential management, and monitoring. Deleting a secret from a public repository does not recall copies an attacker already collected. The credential still has to be rotated, and the access it granted still has to be reviewed.
What businesses should do
Organizations should treat a publicly exposed credential as already compromised:
- Scan repositories continuously for exposed secrets
- Restrict application permissions to what each workload needs
- Rotate compromised credentials immediately
- Monitor unusual cloud activity, including application sign-ins
- Keep response procedures tested, not only written down
The endpoint connection
Cloud identity attacks and endpoint compromise are different threats, and they can meet.
Compromised credentials may let attackers into systems. A compromised endpoint can expose credentials and sensitive files stored on the device.
DriveStrike helps organizations keep control of managed endpoints through remote lock, secure wipe, encryption enforcement, and centralized management. Those controls support data breach protection when a device is lost, stolen, or no longer trusted.
Remote lock, wipe, and encryption sit beside Microsoft Entra security, cloud monitoring, secret management, and credential revocation. They do not replace that work.
Questions and answers
How fast did Microsoft see an exposed credential lead to cloud data?
In one observed case, an AI-assisted operation went from searching public GitHub repositories for credentials to extracting information from Azure Key Vault in about five minutes.
Which Microsoft services were involved?
Microsoft documented activity involving Microsoft Entra ID, Azure resources, application identities, and cloud data repositories, including Azure Key Vault.
Does endpoint control replace cloud identity security?
No. DriveStrike remote lock, secure wipe, encryption enforcement, and centralized management cover managed devices. Credential revocation, secret scanning, and Microsoft Entra controls remain separate work.
The takeaway
When attackers can exploit exposed credentials within minutes, organizations need cloud security and a way to respond quickly on the endpoints that still hold local data.
Learn more about DriveStrike.
Source: Microsoft Threat Intelligence, November 5, 2026.
Start Your Free 30 Day Trial
Exposed credentials can reach cloud data in minutes. Start a free DriveStrike trial and put locate, lock, wipe, and encryption on the endpoints that still hold local copies of company data.