Dr. Bryant Lin shows why AI cannot dismantle medical guardrails

By Elena Rossi · Reporting from Rome ·

On The Free Press, a terminal cancer doctor racing against time warns against dismantling clinical safety rules in the rush to adopt artificial intelligence.

In China, according to Stanford Medicine professor Dr. Bryant Lin, streamlined medical bureaucracy is allowing biotechnology development to outpace the United States. Speaking on the podcast The Free Press with host Maya Sulkin, Lin presented a haunting paradox. He is a doctor and engineer diagnosed with stage 4 lung cancer at age 49. He is racing to harness artificial intelligence to discover life-saving treatments before his clock runs out, while confronting the sluggish reality of modern regulation.

Lin’s predicament exposes a fundamental tension of our era. On one side sits the siren song of technological speed; on the other, the heavy, deliberate architecture of institutional medicine. How much safety regulation should a dying society—or a dying patient—be willing to trade for velocity?

The Temptation of the Fast Track

On the podcast, Lin described a stark economic reality in medical innovation known as Eroom’s Law—the inverse of Moore’s Law. Where microchips double in power while halving in cost, American clinical trials become twice as expensive and less successful over time. Personalized therapies, such as experimental cancer vaccines, can carry price tags between $300,000 and $500,000, leaving patients dependent on trial lotteries or prohibitive private wealth.

The temptation to dismantle regulatory friction is immense. Proponents of rapid deregulation argue that when facing terminal illnesses, bureaucratic delay is itself a lethal factor. If China can accelerate clinical trials by stripping away oversight layers, why should Western medicine remain shackled to institutional review boards and multi-year trials?

Yet this argument mistakes load-bearing safety for mere red tape. The historical record offers a stern warning. Before regulatory tightening, the market was flooded with unproven pharmaceuticals sold without rigorous evidence of efficacy or safety. As Lin noted on the show, AI can suggest millions of biological hypotheses, but it cannot shorten the time required for a human body to respond to a drug. Automated insights still require empirical verification. Without strict trial standards, we risk returning to a wild west of digital snake oil, where algorithmic hallucinations dictate human dosages.

Biological Friction and Institutional Trust

This tension between automated speed and clinical caution plays out daily in patient care. Lin recounted running an open-source AI model on his own home server to analyze his tumor’s DNA sequencing. The software matched his oncologist's treatment recommendation. Sulkin shared her own story of using an algorithm to correctly identify chilblains after a doctor misdiagnosed her with gout.

These anecdotes illustrate the genuine promise of digital tools to educate patients and catch human oversight. But they also reveal the danger of confusing diagnostic software with therapeutic care. Lin cited a case reported by The New York Times. A young woman died after tapering off her seizure medication at the prompting of an AI. Algorithms lack clinical intuition, ethical accountability, and the capacity to build trust. When Sulkin asked whether physicians resist these tools out of professional pride, Lin emphasized that medicine remains a human practice of care, particularly at the end of life.

Democracies must reject both the techno-utopian impulse to break regulatory guardrails and the bureaucratic inertia that leaves life-saving therapies locked in paperwork. Institutions are slow because they are designed to protect human lives from untested haste. The imperative for public policy is not to abandon clinical standards to match authoritarian speed. Instead, we must modernize load-bearing oversight so that genuine science reaches the bedside before time runs out.

Sources

  1. I’m a Doctor Using AI Against My Cancer