When can pharma trust an AI-discovered drug?
AI can design a drug candidate in under two years, yet not one AI-discovered drug has FDA approval. The question is not whether AI can find molecules, but which parts of development you can actually hand over to it.
AI compresses discovery, not proof
Crypto Briefing reports that Insilico Medicine's rentosertib, described as the first oral TNIK inhibitor to reach clinical trials, was designed in roughly 18 months and has entered Phase III testing in China. That speed is real, but Phase III exists precisely to test whether promise holds in humans. TechCrunch notes that AI has not yet produced any FDA-approved drug. So AI compresses the front of the pipeline while the expensive, slow proof stage stays untouched.
- Crypto Briefing says rentosertib also targets cellular senescence, a mechanism beyond its lung disease indication.
- TechCrunch reports Enveda raised $311M to move AI-found, nature-derived drugs into trials.
Agents can run the company, not the clinic
Designing a trial is not the same as running one in human bodies. Stanford Medicine reports James Zou's virtual biotech: tens of thousands of AI agents, no lab, no payroll, covering everything from target discovery to designing clinical trials. That is a striking cost shift for early research. But designing a trial is not the same as running one in human bodies. The hard, physical constraints of testing patients remain outside the software.
- Zou's team built the virtual company on top of a virtual lab that launched in 2025.
- The agents mirror the org chart of an established biotech, per Stanford Medicine.
Cheaper discovery invites more bets, more failures
CNA reports that a small team at Indian startup Peptris is using AI to run many drug programmes at once. The same article quotes experts saying AI addresses only part of a lengthy, risky development process. Lower cost per candidate means you can take more shots, but each shot still faces the same human-trial failure rates. Volume without better prediction just moves the bottleneck downstream.
What this does not solve
None of these sources show an AI-discovered drug that has cleared approval. TechCrunch states this plainly, and Enveda's candidates are still in patient testing. AI also does not remove biology's surprises: Insilico's trial must still prove its senescence mechanism in people. If you treat AI discovery as validated medicine rather than a faster hypothesis generator, you are trusting a story, not evidence.
If AI cut your discovery cost tenfold, would your organization's bottleneck move to trials, and are you staffed for that?
Which decisions about your drug candidates rest on model output alone instead of clinical validation data?
List the three most expensive validation steps in your pipeline and mark which ones AI could actually touch.
Sources
Recommended for you
- Health
Why does FDA approval still leave medical AI unsafe?
Regulators clear AI devices fast, then lose sight of them. Approval was the easy part; the hard part starts the day the device meets real patients.
- Health
What do mental health chatbots actually solve?
Millions now talk to chatbots before therapists. The evidence says they help with access and first-step support, but fail exactly where care gets hard.
- Health
Should patients see what the AI heard in their visit?
A court says ambient AI recordings are not medical records, even as vendors push AI deeper into clinical decisions.
- Health
How do you kill a failing drug before it costs billions?
Novartis lost $30 billion in market value when a late-stage trial missed. An AI start-up had predicted the failure months earlier using virtual patients.
Enjoyed this article?
Subscribe for new articles. No spam. Unsubscribe anytime.