Anthropic Reports Blocking Attempts to Use Claude for Bioweapons Research

Anthropic says it shut down multiple attempts this year by scientists to use its AI technology for research that could aid the development of biological weapons.

The company described five cases in which users "circumvented controls" or tried to "obfuscate" their research goals to get around safeguards. Some involved users in countries Anthropic bars from its models, including Russia, China, and Iran.

One researcher from an "unsupported region" spent weeks planning avian influenza experiments with Claude, the company said, and safety filters limited the work to its weakest models. Anthropic noted it could not be certain the scientists intended harm, since the same information could also help develop a vaccine.

The accounts were banned, but Anthropic did not name the institutions or countries involved. The company said it hopes sharing the examples will spark industry and government discussion about emerging biological risks.

Source: Ars Technica