❗️Anthropic says it blocked five attempts to exploit AI models for potential bioweapons development Artificial intelligence firm Anthropic revealed Thursday that it blocked five separate attempts to misuse its Claude models for activities that could potentially aid in the development of biological weapons. The incidents involved sensitive inquiries regarding chikungunya, avian influenza, and smallpox, including a case where a military-affiliated researcher sought assistance that raised concerns about generating deadlier viral variants. The company's report also detailed attempts to bypass platform safeguards using fraudulent accounts and virtual private networks, alongside separate queries from Yemen concerning guided rocket development. While Anthropic cautioned that detecting such capabilities does not prove malicious intent or a guarantee of success, the findings highlight mounting security risks as advanced AI systems lower technical barriers for specialized research.