JUNE 5, 2026
Ginkgo Bioworks and others build autonomous labs where AI and robots run experiments, raising both efficiency and biosecurity questions
Ginkgo Bioworks, a Boston-based biotechnology company co-founded by MIT graduate students, operates an autonomous laboratory where robots and artificial intelligence handle laboratory tasks including pipetting, experiment scheduling, and protein synthesis. A collaboration between Ginkgo and OpenAI using ChatGPT to design protein experiments resulted in a reported 40 percent reduction in costs and more than 30,000 experiments run over six months, though the paper describing those results has not yet been peer reviewed. Experts including Stanford bioengineering researcher Drew Endy have raised concerns that AI-enabled automation could lower the barriers to misuse, including potential biosecurity threats.
Ginkgo Bioworks, founded roughly two decades ago by four MIT graduate students, was built on the premise that programming cells would eventually matter more than programming computers. Co-founder and CEO Jason Kelly describes early days of little investor interest — "We were living on ramen, buying equipment on eBay" — before the artificial intelligence boom transformed the funding landscape and the company's capabilities.
Today the company operates an autonomous laboratory in Boston overlooking the harbor. Robots resembling single-armed machines, each encased in glass, carry out experiments across pharmaceutical, agricultural, and government projects. A color-coded display tracks each robot's schedule, and an oversized track system moves equipment between stations. Current assignments include engineering microbes for fertilizer improvement and creating proteins capable of producing snow or ice.
The most recent step involves asking AI not just to execute experiments but to design them. Co-founder and COO Reshma Shetty described a collaboration with OpenAI in which ChatGPT was given the goal of synthesizing a specific protein — a task that had previously been the exclusive domain of human scientists. Shetty said the bot performed beyond expectations. The reported outcome was a 40 percent reduction in costs and more than 30,000 experiments completed in six months. Both Shetty and Kelly said humans remain essential for setting the right questions and constraints, but Shetty said the shift has already changed her day-to-day practice: she now spends more time on experimental design and lets robots carry out execution overnight.