Safety Fears as Scientists Make First Viruses Designed by AI

As scientists achieve a breakthrough in designing viruses using artificial intelligence, concerns over safety and security are emerging. The development is a testament to human ingenuity and technological advancement, but it also raises important questions about governance and oversight.

The use of genome language models to design functioning genomes for bacteriophages marks a significant milestone in phage therapy. These viruses, which only infect bacteria, have the potential to revolutionize treatment options for patients with persistent infections. The ability to rapidly design and tune genomes for specific bacterial strains could transform this field and expand biotechnological toolkits.

However, researchers involved in this project are right to highlight the need for caution. The same technology that enables us to design viruses could also be used to create more harmful pathogens. This concern goes beyond the technical capabilities of AI; it speaks to our ability – or lack thereof – to regulate and govern emerging technologies.

The genetic code for viruses that can infect plants, humans, or other animals was intentionally excluded from the AI’s training data to reduce the risk of designing dangerous viruses. However, this raises questions about the limits of such exclusions. Can we truly trust that an AI will only generate genomes for harmless viruses? Or is it more a matter of luck and circumstance?

The issue extends beyond the technology itself to our capacity as a society to handle its implications. The researchers’ call for consultation with safety and security professionals throughout the project highlights a broader problem: we are struggling to establish clear guidelines and regulations for the development and deployment of AI in life sciences.

Some experts argue that the threat from full AI design and writing of a genome is overblown, pointing out that making gain-of-function changes to existing pathogens is a more significant concern. However, this perspective overlooks the very real risks associated with AI-designed viruses. Even if we can contain the harm caused by such pathogens, there are still questions about accountability and liability.

The development of AI-designed viruses highlights the need for a comprehensive approach to governance. We cannot focus solely on regulating the technology itself but must also consider the broader ecosystem in which it operates. This includes safeguards around model development and access, responsible research review, synthesis screening, and established laboratory biosafety and biosecurity protocols.

Ultimately, this is not just about science versus safety; it’s about our capacity to navigate the complexities of emerging technologies and ensure that they serve human interests rather than undermine them. As we move forward with AI-designed viruses, we must be willing to confront the risks head-on and work towards a more robust and inclusive regulatory framework.

The promise of phage therapy is too great to ignore, but so are the potential dangers. We would do well to remember that scientific progress is not a zero-sum game where one side’s gain is the other’s loss. The future of AI-designed viruses will depend on our ability to balance innovation with caution and to prioritize human well-being above all else.