Anthropic announced that it had caught scientists who were trying to develop a biological weapon in “Cloud”

Anthropic’s comprehensive report, which reviews the last eight months, details a long series of misuses of its artificial intelligence model, Claude. Among other things, the use of entities linked to the governments of China and Iran to monitor regime opponents was revealed, as well as attempts by the state media in Russia to spread propaganda masquerading as independent reports around the elections in Moldova. In addition, attempts were documented to use the software to develop weapons Conventional in China, Russia and Yemen, where the context clearly points to the Houthi rebels.

However, the most disturbing threat that emerges from the report concerns biological research. Although the company avoided revealing the identity of the scientists or their countries of origin, it stated that the researchers bypassed geographic protection measures and tried to obscure their true goals. In one case from last May, a researcher used the system to draft a grant application for research into the chikungunya virus, which is transmitted by mosquitoes. The scientist wanted to engineer mutations that will worsen the damage of the virus, and Anthropic emphasized that the main concern stemmed from the fact that the research was intended to be carried out in a military research institute.