AI systems are now capable of designing novel biological sequences, including viruses, raising new questions about biosecurity and the governance of scientific tools. The technology, which can generate proteins and genetic material on demand, is moving faster than the safeguards meant to contain it.
Researchers and policymakers are increasingly concerned that the same models used to accelerate drug discovery and vaccine development could be misused to create pathogens. The risk is no longer hypothetical - several AI models have already demonstrated the ability to propose viable viral sequences that do not exist in nature.
What the technology can do
Generative AI trained on biological data can produce new protein structures and genetic sequences. In the context of virology, this means the technology can suggest modifications to existing viruses or generate entirely new ones. Some of these designs have been shown to be functional when synthesized in a laboratory.
The capability stems from the same underlying architecture used for language models. Instead of predicting the next word in a sentence, these systems predict the next amino acid or nucleotide in a biological sequence. The result is a tool that can explore a vast space of possible biological designs far more quickly than traditional methods.
This has obvious benefits for medicine. Researchers are using AI to design antibodies, engineer enzymes, and identify potential drug targets. But the dual-use nature of the technology means the same tools that help scientists could also help someone with malicious intent.
The security gap
Current screening protocols for DNA synthesis are inconsistent. Many commercial gene-synthesis companies check orders against a list of known pathogens, but the list is incomplete and the checks are not universal. AI-generated sequences that do not match any known threat could slip through.
Governments have begun to respond. The Biden administration issued an executive order on AI that included provisions for biological screening. International discussions are underway at forums like the United Nations and the Biological Weapons Convention. But the pace of policy development lags the pace of the technology.
There is also a question of how to regulate open-source models. Once a model is released publicly, it cannot be recalled. Researchers have debated whether certain capabilities should be kept behind access controls, but the culture of open science pushes in the opposite direction.
What researchers are watching
The most immediate concern is the barrier to entry. As AI tools become more capable and easier to use, the level of expertise required to design a dangerous pathogen drops. What once required years of specialized training could become accessible to someone with a laptop and an internet connection.
At the same time, the same technologies are being used to build defenses. AI is helping researchers identify vulnerabilities in viruses, design broad-spectrum antivirals, and improve surveillance systems that detect outbreaks earlier. The outcome depends largely on how the tools are deployed and who has access to them.
The scientific community is also debating whether to publish research that demonstrates these risks. Some argue that transparency is essential for developing countermeasures. Others say that publishing detailed methods for creating dangerous viruses is irresponsible.
Why this matters for science and research professionals
For working scientists, the practical implication is that AI is now a biosecurity issue, not just a productivity tool. Understanding the risks of AI-generated biological sequences is becoming part of responsible research practice, particularly for those working in virology, synthetic biology, or drug development.
Researchers should be aware of the screening protocols at their institutions and in their supply chains. They should also be prepared to participate in policy discussions about how these tools should be governed. The scientific community's response to this challenge will shape both public trust and regulatory oversight for years to come.
Professionals looking to build skills in this area may find value in AI for Science & Research resources, which cover both the capabilities and the ethical considerations of applying AI to biological problems. For those seeking structured development, an AI Learning Path for Research Scientists can provide a framework for understanding both the opportunities and the risks.
Your membership also unlocks:
