AI tools that aid vaccine design could also accelerate creation of bioweapons, warns former White House Covid adviser

Ashish K. Jha, a former White House Covid‑19 response coordinator and senior fellow at Harvard Kennedy School, cautions that artificial intelligence is lowering the barriers that kept biological weapons rare. He notes that the same generative models used to design vaccines or cancer drugs can be repurposed to create more dangerous pathogens.

Earlier this month Anthropic disclosed five instances where its AI models were employed in work that could support bioweapon development. The cases included assistance in drafting a gain‑of‑function research proposal for a mosquito‑borne virus at a military institute and planning experiments to make bird flu more transmissible to mammals. Anthropic blocked the accounts involved but could not determine the users' intent.

Jha emphasizes that the risk lies not in people explicitly asking AI to build a weapon, but in the technology being used under the guise of legitimate scientific research. The ambiguity makes it difficult for AI providers to police misuse effectively.

He argues that defenses against AI‑enabled biothreats will take years to develop, a timeline that may be insufficient given the accelerating pace of AI capabilities.