Anthropic afirma que ha bloqueado varios intentos de crear armas biológicas con su IA

Anthropic, a leading artificial intelligence (AI) company, has warned that it has thwarted several attempts to exploit its AI models for creating biological weapons, according to a report released on Thursday. In the report, the company detailed how it detected and blocked these illicit uses throughout the year, although it was unable to ascertain whether the studies were for legitimate or malicious purposes, so it opted to block them due to the risk of its models contributing to the development of dangerous pathogens.

While the assessments are valuable in providing evidence of capability, they cannot definitively prove that such capability was ever used to develop biological weapons in the real world, the company explained in the document. As of now, Anthropic has denied having any knowledge of private companies publicly sharing evidence of potential misuse of its platforms for the development of biological weapons.

In an accompanying statement, Anthropic also provided a comprehensive list of cases detected over the past eight months where its AI models were misused, including the chatbot Claude, in China, Russia, and Yemen.

This report comes amid growing concerns among technology experts over the dual-use nature of AI, as models capable of aiding biological research could also be leveraged for hazardous activities, just as there are voices warning of the dangers of increasing autonomy in these models. The tech company stated that it has developed detection systems to identify and block potentially risky uses of its AI models.