Anthropic blocks accounts of several scientists after detecting research with potential to develop biological weapons
In one case, detected in May 2026, a security system blocked a request to use Claude in drafting a funding proposal for gain-of-function research on the chikungunya virus, focusing on its transmissibility and ability to evade the immune system.
Anthropic file photo
Anthropic reported Thursday that it detected and blocked accounts linked to scientists who used its Claude artificial intelligence model in research that could support the development of biological weapons.
The company cited five cases of potential misuse of its models in biological research. Among them, it identified work on viral adaptation, gain-of-function research, and the computational redesign of toxins. Anthropic clarified that it is not claiming the scientists involved intended to cause harm.
Technology
Brussels gains access to Mythos 5, Anthropic's most sensitive AI, after months of negotiations
Carlos Dominguez
In one of the cases, detected in May 2026, a security system blocked a request to use Claude in drafting a funding proposal for gain-of-function research on the chikungunya virus, focusing on its transmissibility and ability to evade the immune system.
The challenge of biology, according to Anthropic
The company also incorporated the information into its security systems to detect and prevent future misuse.
The report warns that biology poses a particular challenge for AI security because the same information can have both beneficial applications, such as vaccine development, and potentially harmful one.
LATEST
