AI Drug Design Faces Biology Limits Despite Investment Surge
Pharmaceutical executives say AI still struggles to predict human biology and drug safety, even as drugmakers expand partnerships with AI firms.
Updated on Nov. 9, 2026 in Biotech
Live Poll
Do you think AI will make new medicines safer and more effective for patients?
Global pharmaceutical CEOs emphasized the limitations of AI in predicting human biology and drug safety, yet drugmakers continued expanding collaborations with AI companies. No AI-designed drug has yet received approval from the U.S. Food and Drug Administration.
Why it matters
Drug development is slow and costly, and AI is widely expected to shorten timelines, but unreliable predictions of off-target effects leave safety as a major hurdle. Approximately one-fourth of late-stage clinical trial failures stem from safety issues.
Venture capital and private equity firms invested $8.4 billion in AI-driven drug development worldwide in 2025, more than double the $4 billion invested two years earlier. Roughly one-fourth of late-stage clinical trial failures stem from safety issues, and it remains unresolved whether AI can reliably predict those off-target effects.
The players
Google DeepMind
Google's artificial intelligence research lab, which developed the AlphaFold protein structure prediction system.
Eli Lilly
A major U.S. pharmaceutical company that announced a five-year, $1 billion AI drug discovery investment with NVIDIA in January.
NVIDIA
A U.S. technology company known for graphics processors widely used in AI, partnering with Eli Lilly on drug discovery and robotics research.
AstraZeneca
A British-Swedish multinational pharmaceutical company that partnered with Insitro to pursue cardiovascular and weight-loss treatments.
U.S. Food and Drug Administration
The U.S. federal agency responsible for approving drugs, and no AI-designed drug has yet received its approval.
The details
AI can design new molecular structures by learning from vast datasets, but current models cannot reliably predict off-target drug effects, and human biology's complexity makes molecular reactions inside the body difficult for AI to forecast. MDR-001 by Mindrank entered Phase 3 trials in February 2026, and Rentosertib by Insilico Medicine followed in July 2026.
Timeline
2018: Google DeepMind unveiled AlphaFold protein structure AI.
2024: AlphaFold3 expanded to predict interactions between DNA, RNA, and other biomolecules.
2025: $8.4 billion was invested in AI drug development worldwide.
February 2026: MDR-001 entered Phase 3 clinical trials.
July 2026: Rentosertib entered Phase 3 clinical trials.
Big Picture
The story extends the 2018 unveiling of AlphaFold, which showed AI could predict protein structures, into the drug pipeline, where that promise is now being tested against approvals and clinical safety. Investment has grown rapidly since, yet no AI-designed drug has cleared FDA approval.
Patients may not see AI-designed medicines soon, since Phase 3 candidates still face years of trials before reaching pharmacies. Consumers should expect development timelines to remain long even as AI tools expand.
The takeaway
AI can speed molecule design, but predicting how drugs behave in human bodies remains the key barrier, so progress will be judged by safe approvals rather than investment totals.
What happens next
Phase 3 trials for MDR-001 and Rentosertib are underway, with commercialization of these candidates still years away.
Further reading
For broader context on drug discovery research, explore the Biotech section.
Source note: This article includes information reported by 조선일보.
Live Poll
Do you think AI will make new medicines safer and more effective for patients?
