Key facts
- Anthropic disrupted malicious uses of its Claude AI models over eight months.
- A suspected Russia-linked cyber espionage campaign used AI to orchestrate and execute cyberattacks.
- Chinese AI firms allegedly attempted to extract and replicate Claude's capabilities.
- A Russia-linked group used AI to build a system that automatically rewrote malware code to evade security detection.
- Anthropic identified new threat actors using Claude to develop software for conventional weapons and intelligence gathering.
- A Yemen-based cell attempted to use Claude for ballistic missile software development.
- A consultant in Mali used Claude to build a mass-surveillance platform monitoring 25 million SIM cards.
Anthropic announced on September 10, 2026, that it had disrupted multiple alleged malicious uses of its Claude AI models over the past eight months. The company's threat intelligence report detailed efforts by cybercriminals and state-backed hackers to leverage AI for orchestrating and executing cyberattacks, often with human oversight.
Anthropic reported disrupting attacks from seven China-based labs, naming tech giant Alibaba, Moonshot, and DeepSeek. Alibaba allegedly conducted the largest 'illicit distillation' attack, aiming to extract capabilities from Claude models to enhance its own Qwen models. Anthropic observed over 151 million exchanges attributed to Alibaba between May and July 2026, peaking at nearly 3 million daily exchanges from more than 3,500 fraudulent accounts. Distillation involves training smaller AI models using output from larger ones to reduce development costs.
Moonshot, the creator of the Kimi chatbot, and DeepSeek allegedly routed live customer conversations, which sometimes contained sensitive information, through Claude and used its responses as training data, Anthropic alleged. Additionally, a hacking group consistent with Russia-based threat actor Midnight Blizzard allegedly conducted phishing, hotel Wi-Fi hijacking, and WhatsApp-takeover operations against Ukrainian government, military, and diplomatic targets, utilizing AI at nearly every stage. This group reportedly used AI to develop a system that automatically detected and rewrote malware code to evade security defenses.
Anthropic also identified new categories of threat actors misusing Claude, including those developing software for conventional weapons, such as firearms, missiles, drones, and bombs, along with their targeting and control systems. The report detailed incidents of operators using Claude for weapons design and development, or to support intelligence gathering and procurement related to weapons programs in China, Russia, and Yemen. Jacob Klein, head of threat intelligence at Anthropic, noted that AI models have become more capable over the past year, increasing risks, particularly in optimizing complex systems like drones or missile software.
Separately, a likely Bamako-based independent consultant working with Mali’s state intelligence service used Claude as the primary engineering workforce to build a population-scale domestic surveillance system that monitors roughly 25 million SIM cards across all three of the country’s national mobile operators. The deployed platform ran on-premises using local models, with Claude providing software design and engineering support. It was designed to generate intelligence dossiers on phone numbers without requiring a court order.