Hackers abused Claude to extract secrets from 1.8 million Android apps

The company said it disrupted several activities linked to the Shiny. Hunters collective, which has been associated with major data theft campaigns. In one operation, an alleged French-speaking Shiny.

Hunters member using the handle “frkoo” deployed a credential-harvesting pipeline across 10 Amazon Web Services EC2 workers. The system downloaded and analysed 1.8 million Android application packages (APKs) from multiple app stores and scanned them for hardcoded secrets using Truffle.

Hog., verified findings were sent in real time to a Telegram group organised into more than 100 source categories. The same actor used another automated process to collect email addresses linked to Git. Hub organisations and obtain Git.

Hub Personal Access Tokens. These credentials were then used to gain initial access in several confirmed breaches. Anthropic said “frkoo” also established a carding operation impersonating French national police to sell stolen payment-card information and victim data.

Suspected Shiny. Hunters members also stole AI API keys and used them to compromise other organisations or conduct reconnaissance. In one case, attackers breached a software-as-a-service provider and stole data belonging to around 200 downstream customers.

AI-powered attacks accelerate. Anthropic said Claude helped one suspected Shiny. Hunters actor extract authentication data and obtain more than 2,100 Azure AD authentication tokens associated with over 40 corporate Microsoft tenants in about 34 hours.

The company said AI agents performed nearly all of the work. Other attacks attributed to Shiny. Hunters affiliates included the theft of 1TB of data from a technology provider, the compromise of an airline and access to an energy company’s systems.

In one case, attackers moved from initial access to bulk data theft within hours, while another actor reportedly progressed from a single stolen developer token to full administrative control in less than three hours. Russian and Chinese espionage activity.

Anthropic also highlighted activity attributed to the Russian espionage group Midnight Blizzard, which allegedly used Claude to automate malware development, phishing, infrastructure acquisition, persistence, command-and-control operations and data exfiltration. The group reportedly targeted more than 20 government, defence, diplomatic, intelligence and foreign-policy organisations.

Anthropic said the campaigns involved device-code phishing, Click. Fix attacks, DNS hijacking, Whats. App account takeovers, cloud email theft and malware targeting Windows, Android and iOS devices.

The report also described an operation attributed to a Chinese-speaking group tracked as GTG-10007. Claude was used as an engineering and orchestration layer for intrusion attempts, reconnaissance of government networks, vulnerability research, exploit development, malware creation and intelligence collection.

The group reportedly operated automated vulnerability-research workflows that uncovered previously unknown vulnerabilities in a major security product and produced working exploits targeting several network and security appliances. Its operations targeted around 50 organisations 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 said it disrupted the groups’ use of Claude, banned the relevant accounts and strengthened its safeguards to detect similar misuse more quickly. The company also said it notified authorities, industry partners and affected organisations.