Concerns are mounting among biosecurity experts about the potential for artificial intelligence (AI) to accelerate the development of biological weapons, as recent advances in AI models have enabled them to generate scientific knowledge and assist in designing novel pathogens. Experts warn that while the current risk remains modest, it could intensify as AI capabilities evolve.
In recent months, AI’s role in biological research has come under scrutiny. In April, biosecurity specialists noted the ease with which chatbots could provide guidance on enhancing the virulence of pathogens. By June, an open letter signed by scientists and AI leaders called for stricter controls on DNA synthesis, citing AI’s ability to rapidly design toxins and viruses. August saw reports of AI being used to create synthetic viruses not found in nature. More recently, Anthropic, a leading AI company, disclosed it had blocked multiple attempts by users suspected of seeking information to weaponize pathogens, prompting its CEO Dario Amodei to advocate for a slowdown in AI development and a ban on its use for producing biological weapons.
Despite these developments, some experts caution against overstating AI’s immediate impact. Drew Endy, a synthetic biologist at Stanford University, described the current biosecurity threat from AI chatbots as “modest” but warned that specialized AI trained on biological data could change the landscape. Unlike language models trained on text alone, these new systems analyze genetic and cellular data, potentially enabling them to engineer novel, viable viral genomes.
Historically, biological weapons have posed serious threats for centuries, with early uses dating back to the 14th century. The 20th century saw state programs amassing extensive stockpiles of biological agents, despite international bans such as the 1925 Geneva Protocol and the 1975 Biological Weapons Convention. While biological weapons have caused few known deaths post-World War II, experts like Thomas Inglesby, director of the Johns Hopkins Center for Health Security, emphasize that the risk remains due to advances in DNA synthesis and biotechnology. The ability to synthesize entire viral genomes cheaply and quickly, exemplified by the 2016 recreation of horsepox virus, underscores how technology has lowered barriers to disturbing capabilities.
AI-driven systems can now retrieve and interpret scientific literature at an unprecedented scale, offering potential bad actors detailed technical knowledge and iterative troubleshooting to overcome scientific challenges. This, Inglesby says, could provide far more practical assistance than simple web searches.
AI models like Evo2, created by researchers at Stanford and the Arc Institute, have showcased how machines trained on millions of genomes can generate thousands of new viral genome sequences, some resulting in viable viruses. Although these models have not been trained on human-infecting viruses to avoid misuse, the underlying technology hints at future risks if similar tools are deployed without safeguards.
The recent attempts to access viral and pathogen-related information using AI appear to involve state-backed laboratories, raising concerns about an escalation in government-led bioweapons research. Experts warn this could rekindle a biological arms race reminiscent of Cold War tensions, with countries accusing one another of clandestine programs.
Efforts to mitigate this threat have included U.S. government regulations requiring DNA synthesis companies to screen orders for hazardous genetic sequences. However, administration changes have stalled enforcement, leaving a regulatory gap that experts say needs urgent addressing. Drew Endy advocates for advancing broader biosecurity measures beyond prevention, including developing defenses against engineered and natural epidemics, upgrading public health infrastructure with early detection sensors, and improving responses to airborne pathogens.
As AI technologies continue to develop, experts emphasize that biological threats—while not new—are evolving in complexity and accessibility. Strengthening international cooperation and regulatory frameworks will be crucial in ensuring these advances do not lead to a resurgence in biological warfare or catastrophic outbreaks.