Far From Killing Us, AI Is Saving Lives

 

The next life-saving medicine may be designed by artificial intelligence. Lawmakers and regulators must ensure that nothing they do gets in the way.

Chemists synthesize thousands of compounds, test each one, and discard nearly all of them. The few compounds that survive go into trials that take over a decade. Of those, only about one in eight are approved. Development costs average $2.6 billion, and most approved drugs never recoup what was spent to develop them.

Drug development is one of the riskiest economic ventures. As a result, diseases go untreated, prescriptions are expensive, billions are spent on drugs that fail, and patients wait decades for a cure that may already exist in a library of millions of untested compounds.

Artificial intelligence is now streamlining the process. In some cases it does work no human team could. Left alone, AI will deliver Americans more life-preserving medicines at a lower cost. 

This month, Insilico Medicine dosed the first patient in a Phase III trial of rentosertib, a lung-disease drug. AI identified the target and designed the molecule. Generate Biomedicines' AI-engineered asthma antibody is now in Phase III with 1,600 patients and requires only two injections a year. At the end of 2023, two dozen AI-designed drugs were in human trials. Today, there are more than 170.

Most exciting is AI’s ability to treat undruggable diseases. Generally, a drug works by fitting into a “pocket” on a protein's surface. Many of the proteins that drive cancer and Alzheimer's have no pocket. In many cases, scientists have known for decades what causes a disease, yet simply can't build a molecule to reach it. AlphaFold, the Nobel Prize-winning AI, now maps a protein's shape in minutes instead of years, and generative AI designs a molecule to fit. Alphabet's Isomorphic Labs has nearly $3 billion in partnerships with Eli Lilly and Novartis targeting these “undruggable” diseases.

With AI, early-stage drug discovery timelines are more than halved. Rentosertib went from target discovery to preclinical candidate in eighteen months with fewer than 80 molecules synthesized. A conventional program takes up to four and a half years and tests thousands. 

With AI, fewer candidates fail on safety. A 2024 analysis found AI-designed molecules cleared Phase I at an 80 to 90 percent rate, well above the 40 to 65 percent average. A candidate that fails on a computer costs a few thousand dollars. One that fails in a clinic costs a few hundred million. Every dollar not wasted on a dead end funds the next attempt.

With AI, drug trials get shorter and smaller. Trials account for roughly two-thirds of drug development costs. The FDA is phasing out animal testing in favor of AI toxicity models – a change then Commissioner Makary tied directly to lower drug prices. In Parkinson's and ALS trials, the FDA has approved AI models that predict how a patient would fare on placebo, so fewer volunteers must take one. 

With AI, rare diseases will not be overlooked. At $2.6 billion per drug, a disease that afflicts 50,000 people cannot justify the investment. At a fraction of that cost, it can. Suddenly, rare diseases stop being charity cases and start being viable markets. 

With AI, competition between drug manufacturers will grow. When only a handful of manufacturers can afford a $2.6 billion bet, patients get fewer options. When that bet shrinks, the field opens to more competitors in more therapeutic areas. 

With AI, patients will have access to more drugs, more quickly, at a lower cost – all without a single government mandate.

But government mandates could derail these medical breakthroughs. Recent coordinated campaigns by former employees of frontier AI labs have many politicians calling for a “pause” on AI research and development out of fear that advanced artificial intelligence might "kill all humans.” Ironically, all the evidence of actual AI use indicates that it is saving lives through medical innovation, not taking them.

Despite the evidence, Senator Bernie Sanders (I-Vt.) has introduced a bill to pause advanced AI development and permanently ban theoretical “superintelligence.” Milder proposals include creating a new AI bureaucracy to approve algorithms before they go to market. President Trump has smartly dismissed AI pessimism. However, federal bureaucrats are still conducting ad hoc reviews of the most powerful models before they are commercially available.

Creating a new agency to pre-clear models or requiring ad hoc reviews from the White House essentially creates a second FDA that innovators must slog through to get drugs to market. AI models are rapidly becoming faster, more accurate, and more powerful. Delaying their use for drug discovery will cost lives rather than save them.

For the first time in history, the next cure may be designed rather than stumbled upon. The only thing that can stop it is a government that can't help itself. Lawmakers and regulators must resist that instinct.

Isabelle Marchese is the Director of Health Policy at Americans for Tax Reform. 

James Erwin is the Director of Innovation Policy at Americans for Tax Reform.

https://www.realclearpolicy.com/articles/2026/10/21/far_from_killing_us_ai_is_saving_lives_1212861.html