AI CEOs Warn: Gene Synthesis Poses Bioterror Threat

CEOs from leading artificial intelligence firms have publicly called on members of Congress to enact legislation that would curtail the ability of malicious actors to develop biological weapons through AI technologies. The signatories, including leaders from Google DeepMind, OpenAI, Anthropic, and Microsoft AI, expressed concerns that rapidly advancing AI could significantly lower the historical barriers to obtaining dangerous biological materials.

Gene synthesis, once a complex scientific endeavor, is now automated and commercialized, with numerous companies worldwide offering custom genetic sequences for research and development. Although many providers vet their customers, not all rigorously screen orders for potentially dangerous applications. This oversight became a concern in 2017 when researchers used mail-order DNA to reconstitute a virus, demonstrating a capability that could be applied to more lethal pathogens.

AI tools, particularly large language models, can now assist in designing novel toxins and pathogens. While creating a functional virus from scratch may still require biological expertise, the potential for AI-driven bioterror attacks is a growing worry. Such attacks could cause mass casualties and widespread panic. Experts note that AI can help users identify and even obscure orders for concerning genetic sequences, challenging existing screening protocols.

Regulatory efforts are underway, including federal guidelines for federally funded research and a bipartisan Senate bill aimed at mandating screening for all US gene synthesis providers. However, screening mechanisms are not foolproof; a recent study showed AI protein design tools generating dangerous sequences that evaded company software. Consequently, AI companies themselves are urged to implement user screening for their biology-focused models to prevent misuse.