![\"\"](\"https:\/\/www.newsbeep.com\/uk\/wp-content\/uploads\/2026\/01\/SEI_280792148.jpg\")
<\/p>\n<\/p>\n
Scientific laboratories can be dangerous places<\/p>\n
PeopleImages\/Shutterstock<\/p>\n<\/p>\n
The use of AI models in scientific laboratories risks enabling dangerous experiments that could cause fires or explosions, researchers have warned. Such models offer a convincing illusion of understanding but are susceptible to missing basic and vital safety precautions. In tests of 19 cutting-edge AI models, every single one made potentially deadly mistakes.<\/p>\n
Serious accidents in university labs are rare but certainly not unheard of. In 1997, chemist Karen Wetterhahn<\/a> was killed by dimethylmercury that seeped through her protective gloves; in 2016, an explosion cost one researcher her arm<\/a>; and in 2014, a scientist was partially blinded<\/a>.<\/p>\n Now, AI models are being pressed into service in a variety of industries and fields, including research laboratories where they can be used to design experiments and procedures. AI models designed for niche tasks have been used successfully in a number of scientific fields, such as biology<\/a>, meteorology<\/a> and mathematics<\/a>. But large general-purpose models are prone to making things up and answering questions even when they have no access to data necessary to form a correct response<\/a>. This can be a nuisance if researching holiday destinations or recipes, but potentially fatal if designing a chemistry experiment.<\/p>\n To investigate the risks, Xiangliang Zhang<\/a> at the University of Notre Dame in Indiana and her colleagues created a test called LabSafety Bench that can measure whether an AI model identifies potential hazards and harmful consequences. It includes 765 multiple-choice questions and 404 pictorial laboratory scenarios that may include safety problems.<\/p>\n In multiple-choice tests, some AI models, such as Vicuna, scored almost as low as would be seen with random guesses, while GPT-4o reached as high as 86.55 per cent accuracy and DeepSeek-R1 as high as 84.49 per cent accuracy. When tested with images, some models, such as InstructBlip-7B, scored below 30 per cent accuracy. The team tested 19 cutting-edge large language models (LLMs) and vision language models on LabSafety Bench and found that none scored more than 70 per cent accuracy overall.<\/p>\n Zhang is optimistic about the future of AI in science, even in so-called self-driving laboratories<\/a> where robots work alone, but says models are not yet ready to design experiments. \u201cNow? In a lab? I don\u2019t think so. They were very often trained for general-purpose tasks: rewriting an email, polishing some paper or summarising a paper. They do very well for these kinds of tasks. [But] they don\u2019t have the domain knowledge about these [laboratory] hazards.\u201d<\/p>\n \u201cWe welcome research that helps make AI in science safe and reliable, especially in high-stakes laboratory settings,\u201d says an OpenAI spokesperson, pointing out that the researchers did not test its leading model. \u201cGPT-5.2 is our most capable science model to date, with significantly stronger reasoning, planning, and error-detection than the model discussed in this paper to better support researchers. It\u2019s designed to accelerate scientific work while humans and existing safety systems remain responsible for safety-critical decisions.\u201d<\/p>\n Google, DeepSeek, Meta, Mistral and Anthropic did not respond to a request for comment.<\/p>\n