In 2026, a team from Stanford University and the Arc Institute in California introduced a groundbreaking AI model called Evo2, capable of designing synthetic genomes that do not exist in nature. Published in the journal Nature, the model functions similarly to ChatGPT, which learns the rules of language to generate grammatically correct sentences. However, Evo2 learns the rules of genome organization by analyzing millions of DNA sequences. It identifies recurring genetic patterns that are crucial for the functioning of living organisms and uses that knowledge to create new genetic sequences that follow the major natural rules of biology. Some of the genomes designed by Evo2 have already been tested in laboratories, leading to the creation of more effective bacteriophages—viruses that kill bacteria. According to a study published in Science in August 2026, these synthetic phages are being explored for use in treating antibiotic-resistant infections, autoimmune diseases, and even tackling plastic pollution. The AI model has generated nearly 300 genomes capable of producing viruses effective against Escherichia coli, a common bacterium. This approach of reprogramming microorganisms is not new. In 2010, biologist Craig Venter demonstrated the ability to insert a completely synthetic genome into a bacterium, proving that it was possible to create life from a manufactured genome. John Glass, a professor of synthetic biology at the J. Craig Venter Institute (JCVI), explains that this opened the door to designing cells with specific properties useful in medicine, research, and agriculture. Until recently, identifying essential genetic elements for life was a slow and experimental process. In 2016, the same team created the smallest known living organism, Syn 3.0, by systematically removing parts of a bacterial genome to determine what was essential for life. Evo2 has proven that training AI on millions of genomic sequences can enable the design of genes and even small genomes, such as those of bacteriophages, according to James J. Collins, a professor of biological engineering at MIT, who was not involved in the study. However, he cautions that while the AI's potential is vast, it is still early, and the proposed sequences must be validated in the lab to ensure they work as intended. For viruses, testing is relatively straightforward: the new genome can be inserted into a host cell, allowing the virus to replicate. For more complex organisms, such as bacteria, the challenge is greater, as the genome must be placed into a functional cell without DNA. John Glass and his team found a solution by using an antibiotic to destroy the original genome and then transplanting the new one, allowing the cell to follow the new genetic instructions. Zumra Seidel, a postdoctoral researcher at JCVI, compares this process to installing new software on a computer that has had its existing programs removed. The cell then functions according to the new instructions. This technique, introduced in March 2026, allows researchers to test AI-designed genomes in real-life conditions. These "zombie" cells could become valuable tools for creating biological systems that produce medicines, clean the environment, or even be used for other purposes. However, James J. Collins emphasizes that AI does not significantly increase the risk of creating dangerous microbes. While AI enhances design capabilities, the actual creation and testing of such organisms require specialized labs and significant resources. In January 2026, researchers from New England Biolabs and Yale University introduced a method to transform viruses into versatile tools, akin to a "Swiss army knife." These synthetic bacteriophages are designed to target antibiotic-resistant bacteria specifically. By incorporating additional genetic tools, such as genes that produce fluorescent markers, scientists can more easily track the phages when they reach their targets, which could be useful in detecting bioterrorism in water systems. In December 2025, researchers from the University of Edinburgh demonstrated that Escherichia coli can be modified to convert a component of plastic (PET) into catechol, a molecule used in the production of pesticides and medicines. James J. Collins concludes that AI is enhancing our ability to understand and modify biology, and while it is powerful, it does not require the complete reprogramming of a genome to achieve these results.