Artificial intelligence may help scientists design medicines, but a Reuters report says the field faces a basic obstacle before models can deliver on the biggest promises: biology does not yet have an internet-scale dataset. Text models benefited from enormous existing collections, while researchers still need far more information about how cells behave, respond and change. That shortage makesdata creation, not only model size or computing power, a central challenge for AI-assisted drug discovery and disease research today.
Alex Rives of Biohub told Reuters that current cell datasets measure in the hundreds of millions of cells, while accurate predictive systems may eventually require billions or even trillions. A coalition involving the US government, Google and Meta has committed a reported**$1.8 billion**to building biological datasets. The group expects an initial dataset in about a year, but usable models could still take as long as five years. Those are development targets, not clinical results or guaranteed treatments.
Companies are also testing narrower ways to apply AI in the life sciences. Danaher plans an autonomous research laboratory intended to design, build and test antibodies faster, while Anthropic has established a biology lab where scientists and AI agents work together. Reuters emphasized that model predictions must still be tested on a physical laboratory bench. Virtual trials may reduce time or cost, but they cannot by themselves prove that a medicine is safe or effective for patients in practice today.
The distinction matters because headlines about curing cancer can run far ahead of the underlying evidence. AI could help identify promising targets, simulate experiments and organize complex biological information, yet the path from a dataset to an approved therapy remains uncertain. For technology readers, the near-term story is infrastructure: better cell measurements, responsible data sharing, laboratory validation and tools that let researchers evaluate predictions. The most ambitious AI biology claims should therefore be treated as research possibilities rather than finished medical breakthroughs.
Source:https://www.reuters.com/technology/artificial-intelligence/ai-wants-cure-disease-it-needs-data-first-2026-10-07/. Real-image credits and licenses are recorded in the preparation handoff.
