If AI Can Design Drugs, Why Aren’t They Approved Yet?

General
1

In 2026, AI can predict protein structures, identify drug targets and generate promising molecules. So why aren’t AI-designed drugs already everywhere?

AI can suggest a molecule, but scientists still need to prove its safety, efficacy, pharmacokinetics (ADME), toxicity and clinical benefit through preclinical studies and Phase I–III clinical trials.

The biggest challenge is that a computer model can predict how a molecule might behave but the human body can behave very differently.Regulators also need validated and explainable AI models, especially when their predictions influence clinical decisions.

The future isn’t necessarily AI replacing scientists.

It is AI + computational biology + medicinal chemistry + clinical research working together. AI can accelerate the discovery. Clinical science still has to prove the drug.

What do you think will be the biggest bottleneck AI accuracy, clinical trials, or regulatory approval?

MBH/PS

2

It’s so important to separate early-stage drug design from clinical drug approval. AI can design a promising molecule in record time, but proving it’s safe and effective in real human biology during Phase I–III trials still takes time. We’re getting closer, but human clinical trials can’t be rushed.

3

I think clinical trials will remain the biggest bottleneck. AI can speed up drug discovery and predict promising candidates, but proving safety, efficacy, pharmacokinetics, and long-term effects in humans takes time. AI may make the process faster, but clinical evidence is still essential before approval.