Your daily read on AI, tech and business
RelPulse
Cybercrime Cybersecurity Threats & Breaches

Cold Relic Hackers Breach Second United States Organization

Share

Russian cyber espionage group Cold Relic leveraged stolen Google Workspace data to compromise a second U.S.-based organization. The sophisticated campaign highlights the growing threat of human-centric social engineering combined with cloud identity abuse, according to findings from The Hacker News.

Cybersecurity professional analyzing network security alerts on multiple computer screens in a dark operations room

Key Facts

  • Cold Relic is also tracked as COLDRIVER, UNC4057, Callisto, Gossamer Bear, and Star Blizzard.
  • The threat group used a Super Admin account and MFA fatigue to breach a U.S. entity in late 2025.
  • Google's GTIG reported that attackers pivoted to compromise a second U.S. organization via Entra ID.
  • Google also released a Threat Tracker report detailing AI tools used by state-backed hackers in China, Iran, and North Korea.

Google's Mandiant and Threat Intelligence Group investigated the intrusion at the second U.S. entity, noting that the attackers utilized identities, corporate staff directories, and data harvested during a late-2025 compromise. Cold Relic, which is also known by the names Callisto, Gossamer Bear, Star Blizzard, and UNC4057, frequently targets government officials, defense contractors, and non-governmental organizations across North America and Europe. The threat group relies on sophisticated phishing and malware campaigns to gather strategic intelligence for the Russian government. Once inside a network, the actors register rogue multi-factor authentication devices, passkeys, and backup codes to maintain persistent access and exfiltrate cloud storage repositories, email, and chat archives.

During the initial late-2025 incident, the hackers gained entry to an unspecified organization by utilizing a residential proxy network close to the victim. The attackers obtained control of a Super Admin account through an MFA push notification fatigue attack after the victim accepted a prompt. The actors then registered their own device in the tenant and downloaded the entire cloud storage drive by mass sharing files and direct downloads, The Hacker News reported. To cover their tracks, the intruders queried security logs, viewed and deleted Alert Center notifications, and exported full employee lists via Google Takeout.

Also read: North Korea Denies Role in October 29 Binance Crypto Attack

Even after being ousted from the first organization, the actors continued their operations by impersonating company employees to target internal staff and professional contacts. This groundwork enabled the transition to the second organization, which used Microsoft Entra ID as its federated single sign-on provider for Google Workspace. Posing as an employee from the first target over a secure messaging application, the attacker shared phishing links under the guise of discussing a technical issue. After harvesting credentials and authentication tokens, the actors registered a new multi-factor authentication device in the SSO solution to maintain access and perform a bulk file download.

In a separate recent report, Google's Threat Tracker shed light on how other state-backed actors are evolving their tactics through artificial intelligence, The Record reported. While Russia continues its espionage campaigns, state-sponsored and financially motivated actors in China are rapidly adopting autonomous multi-agent offensive frameworks like Artex, CAIRN, CyberStrikeAI, and PentAGI to automate intrusions. Google researchers observed Chinese contractors using large language models to translate and organize stolen data, while other groups experimented with tools capable of executing multi-stage cyber attacks at machine speed.

Also read: Binance Loses 6.3 Billion Dollars in Massive North Korea Hack

Background

Cold Relic has a long-established history of conducting strategic phishing and credential harvesting operations against high-value targets across Europe and North America. The threat group has continuously refined its methods to bypass enterprise security controls by combining credential theft with credential abuse, multi-factor authentication manipulation, and social engineering within professional networks.

Concurrently, the broader threat landscape has seen state-backed groups from countries including China, Iran, and North Korea incorporate artificial intelligence into their operations. According to The Record, Iranian actor APT42 used large language models like Gemini for reconnaissance, while North Korean hackers deployed AI to create fake recruiter profiles and execute ClickFix attacks against cryptocurrency organizations.

Also read: SonicOS Security Flaw Added to CISA Catalog Following Attacks

Why It Matters

The convergence of stolen corporate directories, advanced cloud identity abuse, and emerging generative artificial intelligence tools presents a formidable challenge for enterprise security teams. Organizations must adopt rigorous monitoring for suspicious multi-factor authentication device registrations and unusual administrative log activity to mitigate the risk of multi-stage intrusions.

Illustration: AI-generated image, not a photograph of the event.

Watch: this story in the RelPulse Daily Brief

Newer Post Older Post
RelPulse
RelPulse