Google and Mandiant attributed intrusions at two unnamed U.S. organizations to COLD RELIC. The group reportedly took data from the first organization’s Google Workspace environment in late 2025, then used information from that breach to impersonate an employee during a 2026 attack on the second. Google’s account, quoted by The Hacker News, describes a sharp difference in reach: the first intrusion involved a Super Admin account, while the second reached files available to one user.
The first intrusion exposed organizational data and identities
The first organization’s Google Workspace environment was reportedly accessed through a compromised Super Admin account after a user accepted an MFA push request. COLD RELIC then registered an authentication device and used passkeys and backup codes to maintain access. The attackers reportedly mass-shared and downloaded files, exported employee and domain-wide data, and changed or deleted Alert Center notifications.
Access reportedly lasted several months. Unrevoked passkeys could persist across password and session resets, making the registered authentication methods relevant to an account’s recovery.
When an administrator account is compromised, a password reset may not be enough: review and revoke unauthorized passkeys and other authentication methods.
How COLD RELIC reportedly reached a second organization
In 2026, the attackers reportedly posed as an employee of the first organization in a secure messaging app and phished someone at a second U.S. organization. Stolen employee and relationship information likely helped make the approach credible, according to Mandiant as reported by GBHackers.
The second organization used Microsoft Entra ID for federated sign-in to Google Workspace. With federated sign-in, an identity provider—in this case, Microsoft Entra ID—handles a user’s authentication for another service. The reported compromise involved an Entra ID account and a newly registered MFA device; attackers then bulk-downloaded files the compromised user could access. That was a narrower scope than the Super Admin access reported in the first intrusion.
Phishing activity was reported in November 2026
GBHackers reported that Google had observed evidence of new COLD RELIC phishing activity as recently as November 2026.