AI Companies Race to Develop Autonomous Scientific Researchers

Key facts

• AI companies justify their work by citing potential for AI-enabled scientific breakthroughs like curing cancer.
• Large language models (LLMs) currently assist scientists by finding studies, drafting articles, and writing code.
• The ambition is to create AI systems that act as full team members or initiate research with minimal human guidance.
• Google DeepMind's AlphaFold won the 2024 Nobel Prize in Chemistry for predicting protein 3D structures.
• OpenAI launched an AI-for-science team in October 2025, calling an autonomous researcher its 'North Star'.
• Anthropic announced Claude features for biological sciences in October 2025.
• OpenAI released GPT‑Rosalind, a specialized scientific model, and Google released an AI co-scientist tool in February 2025.
• AI-for-science systems often use multiple specialized agents, like Google's supervisor, generation, and ranking agents.
• Stanford's AI for Science Lab created a 'virtual lab' of agents that designed antibody fragments for SARS-CoV-2.
• AI agents cannot physically run lab experiments, a limitation addressed by linking LLMs to robotic labs.
• OpenAI connected GPT-5 to Ginkgo Bioworks' automated labs in February 2026 for iterative experiment proposal and interpretation.
• The GPT-5/Ginkgo system ran many experiments and found a recipe reducing a protein's synthesis cost by 40%.
• A Nature study found AI adoption might reduce the diversity of scientific topics investigated.
• AI excels at analyzing existing data, potentially steering scientists toward established research areas.
• Problems less amenable to AI analysis might receive less attention from scientists.
• Integrating AI effectively into science is more than a technical problem; it requires effort to maintain scientific diversity.