Anthropic disrupts AI misuse for biological weapons, Russia-linked cyberattacks

NEW DELHI: Anthropic has disrupted attempts to misuse its Claude AI models for biological weapons development and a suspected Russia-linked cyber espionage campaign targeting Ukraine, according to a Reuters report.

The AI company disclosed the incidents in its latest Threat Intelligence report, which examines malicious activity involving its technology over the past eight months. It also identified alleged hacking and model-distillation attempts by China-based AI labs seeking to extract Claude’s capabilities.

The findings come as concerns grow over the ability of increasingly advanced AI systems to support cyberattacks, weapons development and other harmful activities. Anthropic’s report also adds to wider industry concerns around the potential risks posed by rapidly improving AI models.

Anthropic said it identified five cases in which scientists used Claude in ways that could support biological weapons development.

In one case, a researcher in a region where Anthropic does not offer its services used virtual private server infrastructure to access Claude and spent several weeks using the model to plan experiments involving the adaptation of avian influenza to mammals.

Anthropic does not support access from countries including Russia, China and North Korea. The company said it subsequently banned the accounts involved and incorporated the findings into its model safeguards, enforcement systems and threat intelligence processes.

The company did not disclose the institutions or countries involved, or provide details about the biological agents or research methods.

Anthropic also identified new categories of threat actors using Claude to develop software for conventional weapons, including firearms, missiles, armed drones and bombs, as well as systems used for targeting and control.

The report linked operators in China, Russia and Yemen to activities involving weapons design, development, intelligence gathering and procurement.

Anthropic said improvements in its AI models have increased the risks associated with such misuse, as the systems are becoming more capable of handling complex technical tasks.

Anthropic also found that cybercriminals and state-backed groups were increasingly using AI to automate large parts of cyberattacks, with humans acting more as supervisors than direct operators.

One group allegedly used Claude across phishing, hotel Wi-Fi hijacking and WhatsApp takeover operations targeting Ukrainian government, military and diplomatic officials.

Anthropic said the group’s methods were consistent with Midnight Blizzard, a Russia-based threat actor that the US government has previously linked to Russia’s SVR foreign intelligence service.

The group allegedly used AI to create a system that could detect when its malware was identified by security systems and automatically rewrite the code to evade detection.

Anthropic also said it disrupted activity linked to affiliates of ShinyHunters, a major cybercrime group associated with attacks on large companies globally.

Anthropic said it disrupted attacks involving seven China-based labs, including Alibaba, Moonshot, DeepSeek and Xiaomi.

The company said operators linked to Alibaba carried out what it described as its largest illicit distillation operation, allegedly attempting to extract capabilities from Claude and use them to improve Alibaba’s Qwen AI models.

Anthropic said it observed more than 151 million exchanges linked to Alibaba between May and July 2026. Activity peaked at nearly three million exchanges a day across more than 3,500 accounts that the company identified as fraudulent.

AI distillation involves using outputs from larger and more expensive models to train smaller models, potentially reducing the cost and resources required to develop new AI systems.

Anthropic also alleged that Moonshot and DeepSeek routed live customer conversations through Claude and used the model’s responses as training data. Some of the conversations allegedly contained sensitive information.

The companies named in the report did not immediately respond to requests for comment.

China’s foreign ministry said it was not aware of the report and maintained that AI should be developed for beneficial purposes, while opposing what it described as the distortion of facts and smears against China.

The report highlights the widening security challenge facing AI developers as increasingly capable models move beyond conventional chatbot use and become tools that can potentially automate sophisticated cyber, weapons and intelligence-related activities.