Anthropic Says Threat Groups Used Claude in Cyberattacks, Including Credential Theft From 1.8 Million Android Apps

· Original article ↗

Summary

Anthropic says ShinyHunters-linked actors used Claude-assisted pipelines to scan 1.8 million Android apps for secrets, while Russian- and Chinese-linked groups used the AI in espionage and cyberattacks. Anthropic banned accounts and strengthened safeguards.

Key points

  • Anthropic says a suspected ShinyHunters member used ten AWS EC2 workers to download and decompile 1.8 million Android APKs, scanning them for hardcoded secrets with TruffleHog.
  • Verified secrets were sent to a Telegram group; a separate pipeline gathered GitHub organization email addresses to obtain Personal Access Tokens.
  • Anthropic attributes breaches, data theft, and credential-token harvesting to suspected ShinyHunters actors, including a software provider incident affecting about 200 downstream customers.
  • Anthropic says Russian-linked Midnight Blizzard used Claude for malware development, phishing, persistence, command-and-control, and exfiltration, targeting more than 20 government and related entities.
  • A Chinese-speaking group tracked as GTG-10007 used Claude for espionage, vulnerability research, and exploit development; Anthropic says exploits were used against government organizations.
  • Anthropic says it disrupted the activity, banned accounts, strengthened safeguards, and contacted authorities, partners, and victims.

Article Details

Attack Vectors
  • ShinyHunters affiliates used automated pipelines to download Android APKs, scan them for hardcoded secrets with TruffleHog, and route verified findings to a Telegram group.
  • An actor collected GitHub organization email addresses and used them to obtain GitHub Personal Access Tokens.
  • ShinyHunters used stolen credentials and tokens to breach organizations and obtain Azure AD authentication tokens.
  • Midnight Blizzard used device-code phishing, ClickFix attacks, and DNS hijacking through compromised hotel Wi-Fi providers.
  • GTG-10007 conducted reconnaissance, vulnerability research, and exploit development, then used exploits against government organizations.
Defensive Notes
  • Anthropic says it disrupted the reported activity, banned associated accounts, and strengthened safeguards and detection for future misuse.
  • Anthropic reported contacting authorities, industry partners, and victims.

Indicators of compromise

TypeIndicatorContext
DOMAINpolicenationale[.]ccDomain used by an alleged ShinyHunters member for a carding shop impersonating French national police and selling stolen payment-card records.

MITRE ATT&CK

People

Threat Actors

Vendors

Products

Tools

Countries

Industries

DefenseAnthropic observed Midnight Blizzard targeting over 20 government, defense, diplomatic, intelligence, and foreign-policy entities.DiplomaticAnthropic observed Midnight Blizzard targeting over 20 government, defense, diplomatic, intelligence, and foreign-policy entities.EducationThe group’s operations targeted around 50 organizations across government, education, retail, energy, technology, healthcare, finance, and manufacturing, with confirmed compromises at an education-technology company, aEnergya technology provider and stealing 1TB of data, compromising an airline, and accessing systems of an energy company.financetargeted around 50 organizations across government, education, retail, energy, technology, healthcare, finance, and manufacturing, with confirmed compromises at an education-technology company, a retailer, and aGovernmentAnthropic observed Midnight Blizzard targeting over 20 government, defense, diplomatic, intelligence, and foreign-policy entities.Healthcareoperations targeted around 50 organizations across government, education, retail, energy, technology, healthcare, finance, and manufacturing, with confirmed compromises at an education-technology company, a retailer,intelligenceThe AI company says that between December 2025 and August 2026, it recorded various forms of artificial intelligence misuse, including for cyber and influence operations,  surveillance, scams, development of biologicalManufacturing50 organizations across government, education, retail, energy, technology, healthcare, finance, and manufacturing, with confirmed compromises at an education-technology company, a retailer, and a Southeast AsianRetailThe group’s operations targeted around 50 organizations across government, education, retail, energy, technology, healthcare, finance, and manufacturing, with confirmed compromises at an education-technology company, aTechnologyharmful activity involving Claude and attributed to ShinyHunters affiliates includes breaching a technology provider and stealing 1TB of data, compromising an airline, and accessing systems of an energy company.

Related Articles