The FDA Didn't Deregulate Medical AI. It Moved the Malpractice Risk Onto Your Doctor — and No Court Has Ruled Who Pays.
The FDA's January guidance is being read as a gift to AI companies, and it is — but the part nobody says out loud is where the risk went. By widening the 'clinician in the loop' rule, the agency let generative-AI clinical tools skip premarket review: if a doctor sees the AI's recommendation and acts, the software isn't a regulated device. That's not new deregulation — it's a stretched 2016 Cures Act carve-out. And it's a liability allocation: the vendor is insulated, the malpractice exposure lands on the physician, and no court has ruled who actually pays when the AI is wrong.
The AI 'Doing Science' Is Mostly a Database Lookup. Zero AI-Discovered Drugs Have Been Approved. The Gap Between Those Two Facts Is the Story.
On VirBench, Claude scored ~17% until researchers bolted on a deterministic tool that looks answers up in NCBI — then it hit 99.7%. On NatureBench, the best of ten frontier agents beat published state-of-the-art on just 17.8% of tasks, winning by 'methodological translation,' not invention. Even OpenAI's flagship science model, GPT-Rosalind, calls AlphaFold as an external tool. The verifiable gains in AI-for-science come from the deterministic scaffolding, not the model's reasoning — and the durable moat is whoever owns the data pipeline and regulatory path, not whoever ships the smartest model.
The Benchmark That Made GPT-Rosalind Famous Measures Bioinformatics. Drug Discovery Is a Different Problem.
GPT-Rosalind scored 0.751 on BixBench — a 0.201-point lead over Gemini 3.1 Pro — and launched with Amgen, Moderna, and Novo Nordisk as partners. Coverage treated this as the opening of AI drug discovery at scale. BixBench measures sequencing data processing, statistical analysis, and genomic interpretation — the computational work biologists do before drug discovery begins. No benchmark measuring actual drug discovery milestones (target identification, lead optimization, ADMET prediction) has been published for GPT-Rosalind. No fully AI-discovered drug has cleared Phase III anywhere. The four largest pharma companies by revenue — Pfizer, Roche, J&J, Merck — are not launch partners. And nobody disclosed what Amgen, Moderna, and Novo Nordisk actually agreed to.
GPT-5.4 found a chemistry improvement. The automated lab that made it possible got buried in the headline.
OpenAI and Molecule.one ran 10,080 reactions and found a real yield improvement for Chan-Lam coupling — a 52% relative gain, experimentally validated. The result is genuinely interesting and the quantitative numbers hold. Three caveats are absent from every piece of coverage: the preprint isn't peer-reviewed, the validation was done by the authors themselves, and the raw data won't be released. Also missing: the irreplaceable component in this experiment wasn't GPT-5.4 — it was Molecule.one's Maria Lab automated chemistry platform. A pharma company with an API key and no automated lab cannot replicate this result.
Mayo Clinic Says It Owns the AI Model It's Building With Microsoft. The Contract Terms Say Something Else — But We Don't Know What.
At Microsoft Build 2026 on June 2, Mustafa Suleiman and Mayo Clinic CEO Dr. Farrugia announced a frontier AI model trained on 54 million de-identified patient records — owned by Mayo, distributed globally via Azure Foundry. The headline is 'Mayo owns the model.' The actual IP terms — does Mayo have unconditional weight access outside Azure, does Microsoft hold any license for MAI development, is the distribution Azure-exclusive — are not publicly disclosed. The model does not exist yet; this is an announcement of intent to build. No coverage has asked the two questions that determine whether 'Mayo owns it' is an IP fact or marketing language. No coverage has asked the FDA whether a hospital-owned frontier AI model distributed as a global API is a device.
AI Drug Discovery Startup Chai Raises $130M at $1.3B Valuation, Lands Eli Lilly Deal
Chai Discovery, an AI biotech startup founded by former OpenAI and Facebook researchers, has closed a $130M Series B at a $1.3B valuation and announced a partnership with Eli Lilly. The company builds custom protein-language models to accelerate antibody design, representing a major bet on AI-driven drug development.