HomeTech NewsCybersecurityGTIG Says COLD RELIC Used One US Victim to Hit a Second

GTIG Says COLD RELIC Used One US Victim to Hit a Second

Google Threat Intelligence Group and Mandiant say the Russian group they now track as COLD RELIC used data stolen from one unnamed US organisation in late 2025 to help compromise a second US organisation in 2026.

In a 17 November 2026 blog post, authors at Mandiant and GTIG wrote that COLD RELIC, “formerly tracked as COLDRIVER and UNC4057; also known as Star Blizzard, Callisto, and Gossamer Bear,” “turns a single cloud compromise into a springboard for ongoing third party compromise and espionage.” “Most recently, Mandiant investigated a 2026 compromise at a US-based organization where COLD RELIC directly weaponized data, the corporate staff directory, and identities stolen during a late-2025 compromise, which Mandiant also investigated, at a separate US-based organization.” Neither organisation is named, and the post gives no victim counts or data volumes.

GTIG said COLD RELIC “is a Russian state-sponsored threat group that the US Department of Justice (DOJ) has assessed operates within Russia’s Federal Security Service (FSB).” That FSB assessment is GTIG’s citation of a Justice Department announcement on Russian intelligence spear-phishing. GTIG said it has tracked and disrupted the group’s phishing campaigns since 2015, issues Government-backed attack alerts when it detects the group targeting Google accounts, and “has observed evidence of new phishing operations as recently as November 2026.”

All of the operational detail below is GTIG’s and Mandiant’s account.

Two US cases, linked by stolen context

In the late-2025 case, GTIG said the operators compromised a US organisation’s cloud productivity environment. They authenticated with stolen Super Admin credentials, together with an MFA push-notification fatigue attack, after connecting through a US residential proxy “in close geographical proximity to the victim.” After an MFA request was accepted, the operators registered a previously unseen device, and used passkeys, MFA backup codes and rogue MFA devices to keep access. With Super Admin privileges, GTIG said, they downloaded the organisation’s entire cloud storage drive, exported full employee lists, and started a domain-wide data export through Google Workspace Customer Takeout. They also viewed, modified and deleted Alert Center notifications. Access, GTIG said, was kept “for several months.”

Customer Takeout is an administrative, tenant-wide export tool, distinct from an individual user’s Google Takeout. Those requests, GTIG noted, are logged in Google Workspace Admin audit logs as CUSTOMER_TAKEOUT_CREATED, and administrators can disable Google Takeout for users at the organisational level.

In the 2026 case, Mandiant investigated a second US organisation that used Microsoft Entra ID as the federated single sign-on provider for Google Workspace. GTIG said the operators impersonated an employee of the first victim over an encrypted messaging app and shared phishing links that stole Entra ID credentials and authentication tokens. After signing in with those stolen credentials, they registered a new MFA device and carried out a bulk download of files in the compromised user’s cloud drive. A COLD RELIC redirector associated with that intrusion, GTIG said, had been submitted to VirusTotal in June 2026, with a query parameter that referenced the first victim organisation. The post includes a redacted, LLM-generated depiction of the social-engineering conversation. It is not a screenshot of the actual chat.

GTIG also said that, after the second compromise, the messaging-app profile used in the approach was changed to a new persona, which GTIG believes was likely used to target other organisations in the same professional circles. When access to the first organisation was removed, it said, impersonating email accounts were used to approach both internal staff and associates in the same circles.

The case is a different kind of GTIG paper from today’s earlier post on fully automated vulnerability exploitation. It sits closer to other TSN accounts of Russia-aligned espionage and intrusion work, including Romania’s finding that it stopped a Turla Cereal email attack and ESET’s attribution of a 2025 wipe at a Ukrainian grain firm to two Russian groups.

What defenders can check

GTIG’s hardening notes go beyond the Takeout log. Password changes alone, it said, have in at least one case left the operators in place because existing sessions were not revoked; it recommends revoking sessions, refresh tokens and third-party authorisations, and removing unauthorised passkeys, backup codes or MFA devices, when an account is believed to have been compromised. It also recommends phishing-resistant MFA, monitoring of new passkey and device enrolments on administrative accounts, and treating residential-proxy logons as a detection problem, because the group has used proxies that geolocate near the victim. Those are GTIG’s defensive recommendations, not a finding that any named organisation has applied them.

A second unnamed intrusion is not a public victim list

The sharpest fact in the post is also the narrowest. Mandiant says it investigated two US organisations, and that data, a staff directory and identities taken from the first were used against the second. That is a linked-case finding, not a register of COLD RELIC’s victims, and not a licence to treat every later phishing attempt in the same professional circle as a confirmed follow-on from these two cases. GTIG itself does not name the organisations, does not publish a count of people or files taken, and does not claim to have measured the full downstream reach of the stolen directory.

The FSB assessment should be read with the same care. GTIG attributes that judgement to the Justice Department and is using it to identify a long-tracked cluster now renamed COLD RELIC. The 17 November post is not itself a new charging document, and it does not attach a new indictment to the 2025 or 2026 intrusions it describes. The useful public facts are the ones GTIG has put on the record: a late-2025 Super Admin compromise, months of retained access, a 2026 impersonation of the first victim’s staff, and a group that, GTIG says, was still running phishing operations as recently as this month.

The defender material in the post is similarly bounded. Logging Customer Takeout, giving administrators a way to disable it, and revoking sessions as well as passwords, are concrete steps GTIG is pointing to because of how these two cases unfolded. They are not proof that Takeout was the only path, or that disabling it would have blocked the second intrusion, which GTIG traces to an Entra ID phishing lure rather than to a Takeout export. The article leaves those limits in place.

Sources

Related stories

GTIG Finds First Fully Automated Vulnerability Exploitation is today’s earlier GTIG publication, and is the next read, even though it describes a different technique.

Follow @tsnmediahub on X for breaking AI news as it happens.

Follow on X

Share this story

Latest stories

More in this category

Latest stories

Explore the TSN Atlas: Every topic we cover, mapped. Click a bubble to dive in.