According to OpenAI, the company has introduced a new evaluation framework designed to measure how artificial intelligence can accelerate biological research in wet laboratory settings.
The framework represents a shift toward real-world testing of AI capabilities beyond traditional benchmarks. OpenAI applied this evaluation approach using GPT-5 to optimize a molecular cloning protocol, a common laboratory procedure used in genetic engineering and biotechnology research.
The work examines both the potential benefits and risks associated with AI-assisted experimentation. By testing AI’s ability to improve actual laboratory protocols rather than theoretical problems, OpenAI aims to provide more practical insights into how these systems can support scientific research.
Molecular cloning involves creating copies of specific DNA sequences, and optimizing such protocols typically requires extensive experimental testing and expertise. The evaluation framework allows researchers to quantify whether and how AI systems can meaningfully contribute to improving these processes.
According to OpenAI, this research explores the dual nature of AI assistance in biology—acknowledging both the promise of accelerated scientific discovery and the potential concerns around AI-guided biological experimentation. The company has not yet released specific performance metrics or detailed results from the evaluation.