GSK tested Chai Discovery's AI-designed proteins in its own lab, found binders for every target, and signed on

Chai Discovery said on Friday, Oct. 9, 2026, that it has started a collaboration with GSK after the drugmaker ran Chai's AI-generated protein designs through its own wet-lab tests and found binders to every target it tested. The designs were made zero-shot, without training on the specific targets. GSK will now get access to Chai's platform, including its folding and design models, for discovery work across its pipeline. Financial terms were not disclosed.

Every AI drug-discovery startup has a slide that says its model designs molecules. Far fewer have a big drugmaker say, on the record, that it took those designs into its own lab and they actually stuck to the targets. That is the part that matters here: GSK ran the test, not Chai, and the designs were made cold, with no target-specific training. It is also a quiet shift in who pharma trusts. GSK already trains its own models on its own wet-lab data, and it is still plugging in an outside startup's design engine. The fair caveats are big ones. "Found a binder" is step one of a very long staircase, nobody disclosed the targets or how many there were, and no money figure came with the press release. A protein that sticks in a dish is not a medicine in a patient. Still, if the AI-designed-drug era is real, it will show up exactly like this: one pharma lab after another quietly running the test and then signing up.

On Friday, 9 October 2026, San Francisco-based Chai Discovery announced a collaboration with GSK after a successful technology evaluation. The company release, carried on Business Wire, is titled “Chai Discovery Announces Collaboration with GSK Following Successful Wet-Lab Evaluation.” MarketMinute and FinancialContent both stamp 9 October 2026, 8:01 a.m. Eastern, which is 12:01 UTC. Chai calls itself an AI company building a computer-aided design suite for molecules. Those lines are Chai’s.

Under the collaboration, GSK gets access to Chai’s platform, including its protein folding and design models, to power discovery across GSK’s pipeline. Protein folding means predicting the three-dimensional shape a protein will take. Design models propose new proteins meant to stick to a chosen target. A pipeline, here, is the set of drug programs a company is already working on. Those lines are Chai’s.

In the evaluation, Chai’s models generated designs against GSK targets zero-shot, without target-specific training. Zero-shot means the model made the designs without first being trained on those particular targets. GSK ran internal wet-lab tests on the designs — real experiments on actual proteins, not only computer scores — and found binders to all tested targets. A binder is a protein that sticks to the target. Those results led GSK to expand into a broader collaboration. The release does not say how many targets were tested, or which diseases they belong to. Those lines are Chai’s.

Joshua Meier, co-founder and chief executive of Chai Discovery, said: “We are pleased to expand our work with GSK following its evaluation of Chai’s technology. The evaluation gave both teams an opportunity to assess Chai’s models against a range of therapeutic targets in the lab. We are excited to build on those results with GSK and support discovery efforts across its pipeline.” That statement is on Chai’s release.

Christopher Austin, senior vice president of R&D Technologies at GSK, said: “GSK is excited to be working with Chai Discovery. We evaluated their models across a panel of diverse targets and saw designs with strong binding affinities. Their technology complements our own internal models that we continuously retrain against our own wet-lab data, all helping to enable us to accelerate the discovery of potential new medicines.” Binding affinity is how tightly a design sticks to its target. Those sentences are his, on the same release.

Chai says its latest model, Chai-3, achieves high target success rates and binding affinities, including on challenging targets. It calls the GSK deal the latest in a string of pharma collaborations announced since the start of 2026. Chai’s newsroom lists earlier partnerships with Eli Lilly, Bristol Myers Squibb, and argenx. Those lines are Chai’s. They are not a score from GSK’s lab.

Chai was founded in 2024 by Joshua Meier, Jack Dent, Matthew McPartlon, and Jacques Boitreaud. Backers named on the release include OpenAI, Thrive Capital, Sequoia, Index Ventures, Kleiner Perkins, Menlo Ventures, General Catalyst, Dimension, and Oak HC/FT. Those names are Chai’s.

No deal value, targets, or disease areas were disclosed. A binder in a dish is not a medicine in a patient.

In plain terms, Chai Discovery said on Friday that GSK tested its AI-designed proteins in GSK’s own lab, found binders for every target it tried, and signed a collaboration. The designs were made zero-shot, with no training on those specific targets. GSK now gets access to Chai’s folding and design models for discovery work across its pipeline. Chai-3 is the company’s latest model. Earlier 2026 partners on Chai’s newsroom include Eli Lilly, Bristol Myers Squibb, and argenx. The release does not name a price, a target list, or a disease area.