On September 11, 2026, Anthropic published its latest threat intelligence report, documenting eight months of Claude misuse that include five cases of research related to biological weapons, a Russia-linked espionage campaign against Ukrainian targets, and new attempts by Chinese labs to extract the model's capabilities.
According to the report, Anthropic detected and shut down five cases in which scientists used its Claude models in ways that could support the development of biological weapons. In one of them, a researcher located in a region where Anthropic doesn't offer service used virtual private server (VPS) infrastructure to access Claude and, over several weeks, used it to plan experiments adapting avian flu to mammals.
The report also describes a threat group whose behavior is consistent with the Russian actor known as Midnight Blizzard, linked by the U.S. government to the SVR foreign intelligence service. This group used AI in nearly every stage of phishing operations, hotel Wi-Fi network hijacking, and WhatsApp account takeovers targeting Ukraine's government, military, and diplomatic corps. It compromised at least three hotel network providers, manipulating DNS records to intercept guest traffic and deliver malware to profiles matching Ukrainian diplomatic and military personnel — a technique similar to one Microsoft had documented on July 31, 2026 under the campaign name "CaptiveCrunch."
The report's third front points to several China-based AI labs, which Anthropic accuses of attempting to illicitly extract Claude's capabilities through mass queries designed to train or fine-tune rival models — an "unauthorized distillation" pattern that echoes similar accusations U.S. agencies had already made against Chinese AI firms.
Carlos Montiel is an enterprise AI solutions architect. He implements LLMs, Agents, RAG, and orchestrators for companies across Guatemala and Latin America. Reach out for a consultation.
Contact Carlos Montiel