AI firm Anthropic has announced the suspension of multiple accounts that may have utilized its AI models for the development of biological weapons.

According to a report released by the company, an analysis of users' activity over a 30-day period identified dangerous signs in approximately 35 independent research projects. These projects included attempts to bypass regional restrictions and conceal research purposes. Specific examples cited include experiments that could potentially alter the nature of infectious diseases, such as avian influenza, and research involving new poisons or toxins.

Anthropic stated that it has not been able to confirm whether the participants in these projects were actually planning to cause harm. Furthermore, while the involved parties described themselves as researchers, they did not disclose their institutional affiliations or countries.

The company is calling for the industry to establish a legal and verifiable mechanism to jointly determine the release timing of high-performance models. In its latest model, "Claude Fable 5," the company stated that it has introduced enhanced safety measures, such as restricting access to queries regarding biological research.

Source: