Artificial intelligence has transformed pharmaceutical science more rapidly than any other part of health care. AI can now:

• Identify molecular targets faster than human researchers
• Compress early discovery timelines from years to months
• Simulate how millions of compounds might behave in the body
• Predict toxicity or poor efficacy before a drug is ever synthesized
• Help design personalized therapies built on real-world clinical data

Recent analyses in Drug Discovery Today and Nature Reviews Drug Discovery highlight that AI is shortening discovery cycles at a pace previously thought impossible. The FDA has acknowledged a sharp rise in drug submissions that incorporate AI and machine-learning components across nonclinical, clinical, and postmarketing phases.

But there is a problem almost no one talks about: AI can now discover the drug faster than our health care system can evaluate it, regulate it, pay for it, or deliver it.

The result is a widening innovation-access gap, a space between what science can do and what systems can actually provide. Patients and clinicians live inside that gap every day, even if they don’t have language for it.

We are entering an era where the science is ready, but the system is not.

When AI innovation meets AI denial

Upstream, in pharma and biotech, AI is celebrated. It is used to design molecules, optimize trial protocols, generate “digital twins,” and identify new indications (advances documented extensively in 2024-2025 biomedical research).

Downstream, in payer and utilization management systems, AI is also spreading, but with a very different purpose. There, it is deployed to:

• Score requests for “necessity”
• Compare cases to historical approval patterns
• Predict high-cost utilization
• Auto-deny based on model outputs
• Escalate specific cases using algorithmic rules

A recent FDA discussion paper and draft guidance highlight both the promise and risk of using AI to support regulatory decision-making, noting that regulatory frameworks are still catching up with AI’s rapid adoption.

The result is an uncomfortable irony: The same AI that accelerates the invention of therapies is now being used to deny or delay them.

Case study 1: a breakthrough therapy stopped at the gate

Consider a composite (but increasingly familiar) scenario in oncology: A cancer center participates in a trial where an AI-assisted regimen for a rare tumor subtype shows dramatic improvement in early results. Clinicians begin prescribing it based on emerging evidence.

When prior authorization requests go in, an automated coverage engine denies them as: “Experimental / Not Medically Necessary.”

Because the system has never “seen” this regimen before, it assigns a low appropriateness score. The therapy discovered through cutting-edge AI is denied by another AI system that cannot recognize what innovation looks like.

Clinicians experience this as a clinical injustice. Patients experience it as abandonment. This is the innovation-access gap made visible.

Case study 2: “We can’t approve what we can’t explain.”

Regulators face their own crisis. By 2024-2025, FDA reviewers reported a surge in drug applications with embedded AI components, yet insufficient clarity on how to assess model transparency, reliability, and credibility.

One theme emerged: “We can’t approve what we can’t explain.”

At the same time, the European Medicines Agency has called for “human-centered AI” in the medicinal product lifecycle while acknowledging that regulatory science is struggling to keep pace.

AI may be ready. The regulatory state is not.

Three different AI systems, three different speeds

We now have three parallel AI ecosystems:

• AI in Pharma and Biotech: Rewarded for novelty and speed
• AI in Regulatory Agencies: Limited by public trust, law, and capacity
• AI in Payer Systems: Rewarded for cost containment and denial efficiency

The FDA’s emergence of internal tools like “Elsa,” designed to accelerate reviews, demonstrates regulators’ attempts to keep up, but these efforts remain early and uneven.

Patients (and clinicians) are trapped in turbulence between these systems.

Algorithmic moral injury: when the system knows, but doesn’t move

Clinicians already know the pain of prior authorization delays. But the innovation-access gap introduces a new form of moral distress.

Imagine explaining to a patient: “There is a therapy that exists.” “We have early evidence it can help.” “But the system doesn’t recognize it yet.”

This is more than administrative friction. It is algorithmic moral injury: The science is strong. The need is urgent. The barrier is artificial (and algorithmic).

This mismatch erodes trust, identity, and purpose across the clinical workforce.

The equity risk: who gets the future first?

NIH’s Bridge2AI program and the NIH Strategic Plan for Data Science stress the importance of inclusive, AI-ready biomedical datasets. But payer and regulatory algorithms often rely on older, incomplete, or inequitable datasets, risking a future where innovative therapies become accessible only to:

• Patients with better insurance
• Academic-center proximity
• Advocates with institutional knowledge
• Communities historically advantaged by the system

Without intentional correction, AI will widen, not narrow, the equity gap.

The leadership imperative: closing the innovation-access gap

Closing this gap requires coordinated leadership at every level:

• Pharma: Build access and equity strategies into every AI-driven development program from day one.
• Regulators: Continue advancing transparency and AI-specific guidance while acknowledging the capacity gap.
• Payers: Treat AI-based coverage tools as high-risk systems requiring bias audits, explainability, and patient safety considerations.
• Health Systems: Track denial patterns as clinical risk, not merely cost metrics.
• Legislators: Update laws that assumed human-only review processes.
• Clinicians: Document and escalate delays attributable to AI-driven decisions.

The future no one wants to admit out loud

If left unchecked:

• AI will design therapies faster than regulators can ethically review them.
• Payer algorithms will decide which innovations are “worth it.”
• Clinicians will carry the emotional burden of telling patients the system hasn’t caught up.

This is the cruelest paradox in modern medicine: We will have the science to save lives, but not the systems to deliver it. Innovation without access is not progress. Innovation without equity is not advancement. Innovation without accountability is not leadership.

AI has accelerated the future of medicine. Now leadership must ensure that patients can reach it.

Tiffiny Black is a health care consultant.