SAN FRANCISCO (Diya TV) — Anthropic said Thursday it disrupted attempts to use its Claude artificial intelligence models for cyberattacks, surveillance, fraud, weapons development and biological research with potential military applications.

The company’s September threat-intelligence report covers activity detected between December 2025 and August 2026 involving Claude Haiku, Sonnet and Opus models. Anthropic said it banned accounts tied to the activity and, when appropriate, shared information with authorities and industry partners.

Anthropic detailed five cases in which Claude was used for biological research that the company said could support biological weapons development. The cases included work involving modified viruses, toxins and other dual-use biological research.

Anthropic said it does not know whether the researchers intended harm and withheld their names, institutions and countries. The company said some of the research could also have legitimate medical or scientific uses.

In one case, Anthropic said its safeguards blocked requests connected to research involving modifications to the chikungunya virus. In another, a researcher used weaker Claude models for work involving avian influenza after safety systems prevented access to more capable models.

The report also describes six cases involving conventional weapons programs. Anthropic said actors in China, Russia and Yemen used Claude for activities including weapons-related software development, intelligence gathering and procurement.

In separate cyber cases, Anthropic identified suspected affiliates of the ShinyHunters cybercrime group using Claude during data-theft and extortion operations. The company also described a Russia-linked espionage operation and China-based actors using Claude for cyberattacks and surveillance.

Anthropic said it also disrupted illicit attempts by seven China-based AI labs to extract or reproduce Claude’s capabilities through model distillation. The company identified DeepSeek, Xiaomi and Moonshot among the organizations whose traffic was involved in some of those efforts.

Anthropic said information gathered from the cases has been incorporated into its detection systems and safeguards.