Artificial intelligence company Anthropic said Thursday that it has thwarted several potential plots this year by scientists using its AI models to conduct research that could contribute toward the development of biological weapons.In a new report laying out misuse of its AI models — which the company said is not typical — Anthropic presented five case studies of scientists using the models in ways that could support bioweapon development. In these examples, the company said actors “circumvented controls we impose to prevent users from unsupported regions accessing our modes, and engaged in other efforts to obfuscate the purpose of their research to evade our safeguards.”“When we detected and investigated these cases, we banned the users’ accounts and incorporated our investigative findings into our frontier model safeguards, enforcement and threat intelligence processes to better prevent, detect and disrupt these activities in the future,” Anthropic said in its report.The company withheld the names of the research institutions, their affiliated countries and the specific biological agents involved. Anthropic was unable to determine whether such research served a legitimate or nefarious purpose, because genuine scientific inquiry that can lead to medical breakthroughs can also result in dangerous engineered pathogens.“Sophisticated threat actors are aware that we … are attempting to detect dangerous uses of our models, and they use the dual-use nature of biology to maintain a kind of ‘plausible deniability’ about their research,” the report stated.“Overt malicious intent is, therefore, often evidence that a particular actor is not all that sophisticated,” it continued. “More sophisticated actors can hide their intent, extracting assistance from an AI model in interactions that look plausibly beneficial, but when put in context and analyzed holistically, can provide clear warning signs of misuse. Those are the kinds of interactions that we report here.”“Sophisticated actors are aware that we ... are attempting to detect dangerous uses of our models, and they use the dual-use nature of biology to maintain a kind of 'plausible deniability' about their research.”- AnthropicOne of the cases involved a scientist in May seeking AI chatbot Claude’s assistance in writing a grant application for funding to conduct research related to the painful mosquito-borne chikungunya virus. While such research can potentially lead to vaccine development, Anthropic said the scientist in this instance wanted to engineer mutations to make the virus more harmful as it repeatedly infected live animals.“One of the reasons we were inclined to think this research was less innocuous was that the institutional affiliation associated with the grant was also a cause of concern,” the report said. “Although information within the application suggested that the research was pursued by civilian researchers, it was intended to be performed at a military research institute.”Anthropic’s report also included misuse of its AI models in instances of cyber operations, surveillance and even building of conventional weapons. But while public attention has focused more on AI’s threat of cyberattacks and failure to align with human intentions, experts also express deep concern over the potential for biological misuse.Anthropic’s case studies are “chilling examples of state-sponsored biological weapons developers tapping into the rapidly advancing capabilities” of leading AI models, Council on Strategic Risks senior fellow Andrew Weber told The New York Times.The report comes as multiple Anthropic employees speak out about the safety and existential risks associated with artificial intelligence, after a former researcher with the business warned that AI companies are barreling “straight to self-improving superintelligence and gambling with our lives.”Relatedartificial intelligenceAnthropic