Table of Contents
Introduction
Anthropic has warned that cybercriminals and state-sponsored hackers alike are using its Claude models for cyber attacks, weapons design, propaganda, and mass surveillance between December 2025 and August 2026.
The threat actors — which the AI company has branded Generative Threat Groups (GTGs) — span state-sponsored groups, financially motivated criminals, commercial spyware vendors, state propaganda institutions, and politically motivated individuals.
"The cybersecurity skills of AI models means that AI has collapsed the labor and tooling gap that used to separate well-resourced, state-sponsored operations from individual operators." — Anthropic
What Are Generative Threat Groups?
Anthropic's 154-page report identifies a wide range of threat actors using Claude for malicious purposes. The use of AI went beyond simple questions and responses — it involved multi-agent frameworks executing reconnaissance, exploitation, and data exfiltration.
Three levels of AI misuse observed:
- Conversational use — Claude as an engineering assistant for malware, phishing kits, and surveillance tooling
- Directed operations — Threat actors directing Claude to execute operations with a human making each targeting decision
- Autonomous operations — Multi-agent frameworks conducting reconnaissance, exploitation, and theft against multiple victims in parallel for hours or days
"At the far end, operations ran autonomously, with minimal human input or supervision." — Anthropic
ShinyHunters & 1.8 Million Android Apps
One of the most notable cases involves GTG-50014 (also known as MeowSHA, frkoo, and blazespider) — a French-speaking operator and suspected affiliate of the ShinyHunters collective.
What this threat actor did:
- Ran a distributed credential-harvesting pipeline across 10 AWS EC2 workers
- Mass-downloaded 1.8 million distinct Android APKs from multiple app-store sources
- Scanned them for hard-coded secrets using TruffleHog
- Sent verified findings to a Telegram group organized into over 100 source types
- Set up a carding shop at policenationale[.]cc impersonating the French national police
With the help of Claude AI, it took a suspected ShinyHunters threat actor about 34 hours to extract authentication data and get more than 2,100 sets of Azure AD authentication tokens linked to over 40 corporate Microsoft tenants. "AI agents performed nearly all of the work."
Russian Group Midnight Blizzard
Anthropic identified GTG-20006, a Russian state-sponsored threat actor sharing tactical overlaps with Midnight Blizzard (APT29/Cozy Bear).
What Midnight Blizzard used Claude for:
- Automating malware development
- Research and infrastructure acquisition
- Phishing campaigns
- Persistence mechanisms
- Command-and-control (C2) operations
- Data exfiltration
- A feedback loop that rebuilt malware whenever security products detected it
Attack techniques used:
- Device-code phishing
- ClickFix attacks
- DNS hijacking through compromised hotel Wi-Fi providers
- WhatsApp account takeovers
- Cloud-email theft
- Windows, Android, and iOS malware
Anthropic observed Midnight Blizzard targeting over 20 government, defense, diplomatic, intelligence, and foreign-policy entities.
Chinese Group GTG-10007
Anthropic identified GTG-10007, a Chinese-speaking operator likely based out of Hunan province — some of whom have been identified as undergraduate students at a Chinese university.
What this group used Claude for:
- Intrusion attempts against production systems
- Reconnaissance of foreign-government networks across the Middle East, Europe, and Southeast Asia
- A vulnerability-research and exploit development effort against major endpoint-security products
- Developing an intelligence-collection platform for bulk-harvesting of open-source material aligned with Beijing's priorities
The GTG-10007 espionage group operated autonomous vulnerability-research workflows while human operators were away, uncovering multiple previously unknown vulnerabilities in a major security product. The actor targeted about 50 organizations across education, retail, energy, technology, healthcare, finance, manufacturing, and government sectors globally.
Influence Operations & Surveillance
Anthropic identified and took down numerous influence operations where Claude played the role of a "sub-editor or content creator" to churn out content at scale.
| GTG ID | Description |
|---|---|
| GTG-04001 | Russian-speaking actor in Bangui running foreign information manipulation in the Central African Republic |
| GTG-54002 | "Influence-as-a-service" operation mass-producing political content across ~70 fabricated news websites (linked to LKM Company, France) |
| GTG-84005 | Commercial election manipulation platform targeting Malaysian users by race and religion |
| GTG-24015 | Used Claude as editorial desk for Sputnik Moldova, RIA Novosti, Sputnik en Español, Sputnik Africa, and RT |
| GTG-34001 | Iranian state-aligned accounts shaping public opinion and disseminating government intelligence |
| GTG-54006 | Automated disinformation network generating fabricated Bengali news and promoting Awami League party |
| GTG-84006 | Influence operation impersonating real activists targeting Iranian audiences (linked to PMOI/MEK) |
| GTG-54004 | Kenyan political content mass-production for pro-administration astroturfing |
| GTG-84002 | Influence operation against the Muslim Brotherhood, Sudan conflict, and UN accountability mechanisms |
| GTG-54009 | Commercial surveillance platform profiling users in Iran and the Persian Gulf region |
| GTG-14010 | China state-aligned operation tracking, profiling, and recruiting Uyghurs in Syria |
| GTG-14020 | Chinese intelligence operation building dossiers on religious leaders and diaspora figures |
| GTG-14021 | Surveillance and transnational repression operations |
| GTG-14022 | China-based "public opinion monitoring" and dissident surveillance operation |
| GTG-34007 | Iranian-nexus actors building government surveillance case-management systems |
| GTG-50027 | Design of national mass interception platform "Lakana 360" for Mali's state intelligence service (25 million SIM cards) |
| GTG-30004 | Iran-nexus OSINT identity-profiling service targeting Israeli and Jewish diaspora organizations |
| GTG-30005 | Iran-nexus gathering data for targeting US naval forces and building mass-surveillance platform |
| GTG-30006 | Iranian threat actor developing malware, phishing portals, and modular Windows implant SECOMS64 |
Weapons Design & Military Applications
Anthropic also neutralized several weapons-related misuse efforts:
- Northern Yemen-based threat actors — developing guided weapons
- Two China-based operations — drafting a specification for an anti-torpedo fire control system and building targeting software for electronic warfare
- Russia-based operation — engineering a full-stack autonomous first-person-view (FPV) kamikaze drone swarm
AI models are being used not just for cyberattacks, but for weapons design and military applications — a significant escalation in the threat landscape.
How Anthropic Responded
Anthropic took several actions in response to the misuse:
- Disrupted the actors' use of Claude for harmful activities
- Banned the threat actors' accounts
- Adjusted guardrails based on observed malicious use
- Added detection measures to identify future misuse faster
- Contacted authorities, industry partners, and victims
"We hope that sharing these early insights with the public helps inform governments, the industry, and the general public on the nature of these risks, and the safeguards that are necessary for ensuring the safe deployment of AI models." — Anthropic
What This Means for Your Business
This report has significant implications for organizations of all sizes:
- AI has democratized advanced cyberattacks — individual operators can now do what previously required state resources
- Attacks are faster — ShinyHunters moved from token theft to admin control in under 3 hours
- AI-powered reconnaissance — Autonomous vulnerability research can uncover unknown bugs
- Surveillance at scale — Nation-states are using AI for mass surveillance capabilities
- Your AI tools can be weaponized — Even legitimate AI tools can be misused if compromised
Organizations should assume that threat actors are using AI to accelerate and scale their attacks. Traditional security approaches may not be sufficient.
Final Thoughts
Anthropic's report reveals an alarming picture of how AI models like Claude are being weaponized by cybercriminals, state-sponsored actors, and influence operations worldwide. The scale, speed, and sophistication of these attacks demonstrate that AI has fundamentally changed the threat landscape.
Key takeaways:
- GTGs are diverse — state-sponsored, criminals, spyware vendors, and individuals
- ShinyHunters — 1.8M Android APKs scanned, 2,100+ Azure AD tokens stolen in 34 hours
- Midnight Blizzard — Automated malware development and C2 operations
- GTG-10007 — Chinese group targeted 50 orgs globally with autonomous exploits
- Weapons design — AI used for guided weapons and drone swarms
- Mass surveillance — Nation-states building AI-powered surveillance platforms
- Defenders must adapt — Traditional security is no longer enough
If your organization needs assistance with AI security, threat hunting, or defending against AI-powered attacks, contact SystemLogicIT for expert guidance.