Anthropic Blocks Accounts After AI Used in Sensitive Bio-Research
Anthropic reported that scientists misused the Claude AI model for sensitive biological research involving potential pathogens and gain-of-function experiments, prompting account bans.
Artificial intelligence company Anthropic has revealed that scientists used the Claude model for sensitive biological research, some of which could potentially be repurposed for biological weapons development. The company blocked accounts that violated its usage policy, but emphasized that it could not determine the researchers' true intentions.
The cases were disclosed in an Anthropic report detailing the misuse of its AI models. Among other activities, Claude was used in research on pathogens with pandemic potential and in gain-of-function experiments, where viruses are modified to examine how their traits might change.
In one instance, a scientist asked for Claude's assistance in writing a grant application for research on the chikungunya virus. The stated research goal was to identify mutations that could affect the virus's infectivity and immune evasion, and subsequently introduce those changes into the virus itself.
Chikungunya is a virus transmitted primarily by mosquitoes that can cause fever and severe pain. Anthropic noted that such research can serve legitimate medical purposes, including the development of vaccines and treatments. However, the company grew especially concerned after discovering that the work was planned to take place at a military research institute.
Anthropic clarified that no evidence was found to prove the researchers intended to weaponize the virus. Nevertheless, the company determined that the activity violated its safety guidelines, blocked system access, and closed the accounts linked to the incidents.
