AI Is rewiring drug discovery. Regulators must keep up

Artificial intelligence, particularly in the form of specialized and domain-trained models, is transforming how we design, test, and validate new therapeutics. Nowhere is this transformation more urgently needed than in the fields of neurology and neuropsychiatry, where treatment breakthroughs have long lagged behind patient needs.

Despite decades of research, the brain remains one of the least understood organs in the human body. Mental health disorders, neurodegenerative diseases, and neurological syndromes continue to carry staggering societal and economic costs. But new computational tools are offering unprecedented insight — and hope. Technologies like Quantitative Structure-Activity Relationship modeling now allow researchers to create highly specific simulations of how chemical compounds interact with biological targets, including human neurons. 

These tools go far beyond traditional pharmacology, giving scientists the ability to predict with fine detail the efficacy, toxicity, and pharmacokinetics of drug candidates before a single dollar is spent on a wet lab experiment.

Adding to this, the rise of large language models (LLMs) and generative AI systems allows us to simulate chemical synthesis pathways in silico — helping researchers rapidly iterate new molecular designs, test binding profiles, and simulate receptor interactions. Taken together, these technologies offer something the industry has long struggled to achieve: speed without sacrificing safety.

The implications are profound. If deployed at scale, these AI-enabled technologies could unlock a flood of new drugs — targeted not just at blockbuster indications like diabetes or oncology, but at rare diseases and underserved mental health conditions long ignored by the traditional pharmaceutical model. But this future is not guaranteed. While the science and software are accelerating, the regulatory and funding models remain stuck in the past.

Pharmaceutical development is still governed by outdated incentive structures. Regulatory frameworks often lag far behind the pace of technological change, and funding remains heavily tilted toward a narrow band of “safe bets” — indications and molecules that venture capital believes will receive timely approval. The result is a crowded landscape of me-too drugs and incremental innovation.

Last week, my company, PharmAla Biotech, took a bold step toward changing that. We launched Phenesafe AI, the world’s first computational value chain designed specifically for the discovery of novel phenethylamine and MDXX-class molecules. These compounds — including the well-known MDMA — hold transformative potential for treating a range of neuropsychiatric conditions. We already know from extensive clinical research, including two published Phase 3 trials, that MDMA-assisted therapy can help people suffering from treatment-resistant PTSD. But we believe this is only the beginning.

The MDXX class of molecules remains vastly underexplored. With the help of Phenesafe AI, we are using computational chemistry and machine learning to explore this chemical space more thoroughly than ever before — testing new analogues, modeling their interactions with receptors, and identifying candidates that could offer superior efficacy, fewer side effects, and more targeted therapeutic action.

But for any of this to help patients, regulators must meet this technological revolution with policy innovation of their own. Drug authorization is just the beginning. Health Canada, to its credit, was one of the first national regulators to permit the use of MDMA through the Special Access Program (SAP). That was two years ago. Yet the number of Canadians who have actually received MDMA-assisted therapy under SAP remains vanishingly small.

Authorization without infrastructure — without delivery mechanisms, provider training, reimbursement pathways and awareness — is a hollow victory. Patients do not benefit from a drug that exists on paper but is functionally inaccessible in practice.

To unlock the full potential of this next generation of neuropsychiatric drugs, regulators must rethink both process and priority. That means streamlining access for compassionate use. It means developing regulatory pathways that account for AI-generated preclinical data. And it means creating incentives that allow funding to follow innovation, not bureaucracy.

The problem today is not that pharma lacks innovation. It’s that innovation is often punished with longer timelines, greater risk, and higher capital requirements. The industry crowds into a small number of clinical indications because those are the conditions that investors believe will secure regulatory approval. The result? Dozens of companies all chasing the same endpoints, while vast swaths of medical need remain unmet.

But technology has changed the game. AI allows us to move faster, design smarter, and fail earlier. If regulators adjust their models to reflect this new technological paradigm, the money will follow. Venture capital is not afraid of risk, but it does fear stagnation and ambiguity. A regulatory system that embraces AI-driven development, that rewards novel mechanisms of action, and that accelerates access for patients will unlock billions in new investment.

Patients, not just shareholders, will be the biggest winners.

This is the beginning of a new era in drug discovery. But without policy to match the pace of progress, innovation will stall at the edge of possibility. We are building the tools to find the drugs of the future. It's time for regulators to build the roads that get those drugs to patients.

---

Nick Kadysh is the Board chair of PsyCan, the trade association for Canadian medical psychedelics companies, and the CEO of PharmAla Biotech.