Anthropic reports disrupting bioweapons attempts, Russia-linked espionage and Chinese "distillation" of Claude
In a threat-intelligence report published on 10 September 2026, Anthropic described misuse of its Claude models over eight months, including 151 million exchanges attributed to Alibaba and a hacking campaign against Ukraine.
On 10 September 2026 Anthropic published a threat-intelligence report covering malicious use of its Claude models over the previous eight months. Reuters reported the main findings the same day.
Three lines of misuse
Biological research. Anthropic documented five cases of scientists attempting to use Claude for research that could support biological-weapons development. In one case a researcher used VPN infrastructure to access the model "for weeks" while planning experiments on adapting avian influenza to mammals.
Cyber espionage against Ukraine. A hacking group that Anthropic links to Midnight Blizzard, a unit associated with Russia's foreign-intelligence service, used the model across phishing, hotel Wi-Fi hijacking and WhatsApp-takeover operations targeting Ukrainian government, military and diplomatic sectors. The group also used AI to build systems that automatically detect and rewrite malware to evade security tools.
"Distillation" by Chinese labs. Anthropic said it disrupted attempts by seven China-based labs to extract Claude's capabilities. It attributed 151 million exchanges between May and July 2026 to Alibaba, peaking at nearly 3 million per day from more than 3,500 accounts it described as fraudulent. Moonshot (creator of the Kimi chatbot) and DeepSeek allegedly routed live customer conversations through Claude and used the responses as training data.
Why it matters
The report is one of the most detailed public accounts so far of how frontier models are being probed by state-linked and commercial actors at the same time. For organisations deploying AI in government or critical infrastructure, it is a reminder that model access controls, account verification and logging are part of the security perimeter, not an afterthought.
The companies named in the report had not publicly responded at the time of the Reuters article.
