In June 2026, a group of eminent mathematicians did something unusual. They issued a declaration.
The Leiden Declaration on Artificial Intelligence and Mathematics set out a clear statement of values for how their field should navigate the age of AI — insisting on transparency, human accountability, and the protection of genuine mathematical understanding in an era when machines can now solve problems that resisted human effort for decades.
It was a serious, coordinated response from a serious discipline.
And according to Yale computational biologist C. Brandon Ogbunu, biology needs to do the same, and fast.
Why Mathematics Got There First
The mathematicians moved quickly for a specific reason: their field was directly disrupted. AI systems have begun solving advanced mathematical problems independently, raising immediate questions about proof verification, credit, and the integrity of the discipline. And mathematics, conveniently, has a structural advantage when it comes to AI oversight: proofs are transparent and independently verifiable. A counterfeit mathematical proof can, in principle, be detected by comparing it to the genuine article.
Biology has no such luxury.
The Forgery Problem Is Harder in Biology
This is the central tension at the heart of Ogbunu’s argument. In mathematics, truth is fixed. A theorem that was valid in 1850 is still valid today. In biology, truth is noisy, contextual, and often temporary. Principles that hold in a petri dish break down in a living body. Effects that appear in mice vanish in primates. A mutation that confers drug resistance in one genetic background may be neutral or even harmful in a slightly different one — a phenomenon called epistasis, which is not an exotic edge case but a fundamental feature of how biological systems work.
When a mathematician worries about AI, they fear a convincing fake they can ultimately disprove. When a biologist worries about AI, they face something more unsettling: they often cannot even define what the authentic version should look like. Biological “truths” are frequently better described as powerful generalisations with well-characterised exceptions. An AI that confidently generates biological conclusions without understanding that context isn’t producing counterfeits — it’s producing plausible-looking outputs that may simply be wrong in ways that are very hard to detect.
What a Biology Declaration Should Demand
Ogbunu argues that a biological equivalent of the Leiden Declaration should borrow its core commitments — mandatory disclosure of AI tool use, human accountability for findings, credit belonging to people rather than systems, and protection for early-career researchers from incentives that reward volume over genuine insight — while adding provisions specific to the life sciences.
These would include rigorous validation of AI-generated hypotheses against real wet lab experiments, careful governance of training data drawn from shared biological resources, and especially heightened scrutiny for any AI output that will eventually inform patient care or affect an ecosystem. The stakes in biology are not abstract. AI’s mistakes and successes in this domain will manifest in living bodies, with all the ethical, emotional, and legal consequences that entails.
The Hardest Requirement: Staying Current
There is a deeper challenge that makes a biology declaration structurally different from the mathematics version. Mathematical truths endure across centuries. Biological understanding shifts — sometimes rapidly. A policy written to match the AI landscape of this summer could be miscalibrated by winter. A declaration intended to last a decade might spend most of that decade playing catch-up.
This means a biology declaration cannot simply be written and signed and left to stand. It needs to be adaptive by design: versioned, dated, revised on a published schedule, and amended openly by the community it represents. Updating it should be treated as a normal, positive act of scientific maintenance — not a sign that the original got something wrong.
Ogbunu notes, with some wryness, that life scientists should be well equipped for exactly this challenge. Biology has always understood that structures unable to adapt to their environment rarely survive. It would be an odd betrayal of the discipline to produce a declaration that forgets its own foundational principle.
The Urgency Is Real
AI is already reshaping biological research at pace. It is generating hypotheses, analysing genomic data, predicting protein structures, and proposing drug candidates — all with a speed and scale no human team can match. That is genuinely useful. It is also genuinely risky, in ways that are specific to biology and that the general AI ethics conversation has not yet fully grappled with.
The mathematicians saw what was coming for their field and chose to articulate their values before the moment of crisis. Biologists have the same opportunity — and arguably more reason to act quickly, given how directly their work touches human health. The declaration doesn’t need to be perfect. It needs to exist.