
Mandiant, part of Google, says the Russian state-backed group COLD RELIC (formerly tracked as COLDRIVER, also known as Star Blizzard, Callisto and Gossamer Bear) is turning one cloud breach into a springboard for the next. In a report dated November 17, Google’s threat intelligence team describes a late-2025 intrusion at one US organization and a 2026 intrusion at a second, where the group used the first victim’s stolen staff directory, contacts and identities to approach its next target. Google notes that the US Department of Justice has assessed the group operates within Russia’s Federal Security Service (FSB), and says it saw new phishing operations as recently as November 2026. It did not name either victim or say what was stolen, The Record notes.
In the first case, COLD RELIC logged in to a Super Admin account with stolen credentials and a multi-factor “push fatigue” attack, using a residential proxy located near the victim to avoid location-based alerts. After one MFA prompt was accepted, it registered a previously unseen device and added passkeys, MFA backup codes and rogue MFA devices to keep access. It downloaded the organization’s whole cloud drive, ran a domain-wide Google Workspace Customer Takeout export, and viewed, changed and deleted Alert Center notifications. It stayed for months, syncing mail through OAuth to external clients and signing back in with attacker-enrolled passkeys that survived password and session resets. After it was removed, it used email accounts impersonating the victim’s employees against colleagues and associates.
In the second case, COLD RELIC messaged a person at another US organization over an encrypted messaging app while impersonating an employee of the first victim, then sent phishing links, during the conversation, that led to documents matching what they were discussing. The links stole Microsoft Entra ID credentials and tokens, and the group registered a new MFA device and downloaded files from the user’s cloud drive. Mandiant says phishing links typically start at a compromised website that fingerprints visitors before redirecting them, and often carry an attacker-set parameter naming a previous victim or lure. The messaging account later changed to a new profile name, which Google believes was a new persona for targeting others in the same circles.
How to check if you’re affected
Affected products are Google Workspace and Microsoft Entra ID tenants, the two environments in Mandiant’s cases, and the people in them who hold admin or sensitive access. Mandiant’s report recommends these checks:
- Admin logs: alert on creation of tenant-wide exports (
CUSTOMER_TAKEOUT_CREATEDin Workspace Admin audit logs), spikes in file downloads or external sharing, and bursts of MFA prompts followed by a successful sign-in from the same IP address. - MFA and passkeys: audit passkey registrations and new MFA devices, especially on admin accounts, and require approval for new authenticators.
- OAuth grants: review third-party app permissions and flag broad mail scopes granted to unexpected clients such as Thunderbird or Proton Mail.
- After a suspected compromise: a password reset alone is not enough. Mandiant says to revoke all active sessions, refresh tokens and OAuth permissions in both Workspace and Entra ID, and remove unauthorized passkeys, backup codes and MFA devices. In one case the victim reset passwords but did not revoke sessions, and the group kept its access.
- Prevention: Mandiant recommends phishing-resistant MFA (FIDO2/WebAuthn hardware keys, not SMS, voice or push), alerts on sign-ins from residential proxy or VPN ranges for privileged accounts, and data-loss-prevention rules that warn or block files matching sensitive patterns (such as credentials) from being uploaded or shared externally.
