New research has been published by OpenAI, examining whether advanced AI models can accelerate biological research within the wet lab, rather than just supporting theoretical science.
Working with biosecurity firm Red Queen Bio, researchers tested GPT-5 within a tightly controlled molecular cloning system designed to measure practical laboratory improvements.
Across multiple experimental rounds, GPT-5 independently proposed protocol modifications, analysed results and refined its approach using experimental feedback.
The model introduced a previously unexplored enzymatic mechanism that combines RecA and gp32 proteins, along with adjustments to reaction timing and temperature, resulting in a 79-fold increase in cloning efficiency compared to the baseline protocol.
OpenAI emphasises that all experiments were carried out under strict biosecurity safeguards and still relied on human scientists to execute laboratory work.
Even so, the findings suggest AI systems could work alongside researchers to reduce costs, accelerate experimentation and improve scientific productivity while informing future safety and governance frameworks.
Would you like to learn more about AI, tech and digital diplomacy? If so, ask our Diplo chatbot!
