"Alarming" warning or overreaction? Report on AI-powered bioweapons divides experts.
Anthropic just published a report acknowledging, for the first time among major AI companies, that its tool may have been used by scientists attempting to develop biological weapons. And the scientific community is nowhere near a consensus on how serious that actually is.
The report describes five cases in which scientists from unnamed countries tried to use Claude for research involving viruses and toxins with the potential to cause harm. The company is clear on one point: it doesn't claim malicious intent was involved. But it flags an uncomfortable detail — when blocked by the model's safety restrictions, these scientists found workarounds.
On one side, heavyweight voices are treating this as a serious warning. Andrew Weber, of the Council on Strategic Risks, called the cases "chilling" examples of state-linked bioweapons developers tapping into rapidly advancing AI capabilities. Ashish Jha, who helped lead the U.S. COVID-19 response, put it even more bluntly: he wrote that we're "sleepwalking into a potentially huge disaster" and that we won't get five more warnings before something goes wrong.
On the other side, not every scientist is buying the alarm. Virologist Kristian Andersen, of Scripps Research, publicly dismissed the way Anthropic framed the cases as an overreaction.
In the end, the report may say less about the real risk of AI-enabled bioterrorism and more about how the scientific community still doesn't know how to calibrate this kind of alarm — in either direction.