Anthropic, the developer of the Claude AI model, has reported that multiple threat groups, including state-sponsored espionage actors linked to Russia and China, have attempted to misuse its AI for malicious purposes. Between December 2025 and August 2026, the company observed various forms of AI misuse, encompassing cyber and influence operations, surveillance, scams, and the development of weapons.
Among the financially motivated groups, Anthropic identified activities linked to the ShinyHunters collective. An alleged French-speaking member, operating under the handle 'frkoo', deployed a credential-harvesting pipeline across ten AWS EC2 workers. This pipeline mass-downloaded 1.8 million distinct Android APKs from multiple app stores, decompiled them, and scanned for hardcoded secrets using TruffleHog. Verified findings were then routed in real-time to a Telegram group organized by over 100 source types.
This same actor also established an automated process to collect GitHub organization email addresses, which were subsequently used to obtain GitHub Personal Access Tokens (PATs). Anthropic states that these two pipelines provided the initial-access credentials 'frkoo' leveraged for the majority of confirmed breaches associated with the hacker. Additionally, 'frkoo' reportedly operated a carding shop, policenationale[.]cc, impersonating the French national police to sell stolen payment card records and full cardholder information.
Suspected ShinyHunters members also engaged in stealing AI API keys, using them for breaching other organizations or for reconnaissance. In one instance, they breached a software-as-a-service provider, compromising data belonging to approximately 200 downstream customers. Anthropic noted the speed of these attacks, citing one case where a suspected ShinyHunters actor, aided by Claude AI, extracted over 2,100 sets of Azure AD authentication tokens linked to more than 40 corporate Microsoft tenants in about 34 hours, with AI agents performing nearly all the work.
Other harmful activities attributed to ShinyHunters affiliates include breaching a technology provider to steal 1TB of data, compromising an airline, and accessing systems of an energy company. The group demonstrated rapid progression from initial access to bulk data theft, in one case moving from a single stolen developer token to full administrative control in under three hours.
Anthropic's report also detailed activity from the Russian espionage group Midnight Blizzard, which utilized Claude to automate malware development, research, infrastructure acquisition, phishing, persistence, command-and-control (C2) operations, and data exfiltration. This group also implemented a feedback loop to rebuild malware upon detection by security products. Midnight Blizzard targeted over 20 government, defense, diplomatic, intelligence, and foreign-policy entities, employing tactics such as device-code phishing, ClickFix attacks, DNS hijacking via compromised hotel Wi-Fi, WhatsApp account takeovers, cloud-email theft, and Windows, Android, and iOS malware, with Claude integrated across all attack stages.
The Russian group's operations were automated through AI-driven workflows built around Claude Code skills, with human operators primarily refining these skills as needed.
Furthermore, Anthropic identified an espionage operation by a Chinese-speaking group tracked as GTG-10007. This group used Claude as the "engineering and orchestration layer of a coordinated offensive program" involving intrusion attempts against production systems, reconnaissance of foreign-government networks across the Middle East, Europe, and Southeast Asia, vulnerability research, exploit development against major endpoint security products, malware development, and the construction of an intelligence-collection platform.
GTG-10007 operated autonomous vulnerability-research workflows, even when human operators were absent, uncovering multiple previously unknown vulnerabilities in a major security product and delivering working exploits for several families of network and security appliances. These exploits were then leveraged against various government organizations globally. The group's operations targeted approximately 50 organizations across government, education, retail, energy, technology, healthcare, finance, and manufacturing, with confirmed compromises at an education-technology company, a retailer, and a Southeast Asian government agency.
Anthropic confirmed that it disrupted the actors' use of Claude for these harmful activities, banned their accounts, adjusted its guardrails to detect future misuse more rapidly, and contacted authorities, industry partners, and victims.
