The SAIR Foundation believes that, as artificial intelligence profoundly reshapes the landscape of scientific discovery, it should remain firmly anchored in the core principles of science: open academic exchange, rigorous peer validation, and deep respect for intellectual contribution.
These principles take practical form in Open Math OS, the operating system for collaborative math research and scientific computing. It brings open frontier models, a scientific compute cloud, and collaboration tools into one system, with the mathematical community in control of its data.
1. Why It Matters to Mathematicians and the World
From Black Box to Community-Owned Infrastructure
Mathematics has always advanced through transparency: every theorem and every proof must be open to intense scrutiny, verification, and extension by peers. SAIR Foundation holds that the future of AI-assisted mathematics—the vision of Math 2.0—requires community-owned infrastructure. Researchers must be able to independently inspect, continually refine, and run the models they rely on within environments they control, rather than becoming chronically dependent on the opaque, closed-source APIs of commercial entities, where the underlying reasoning algorithms remain hidden. The same conviction underlies Open Math OS, which is built on leading open-weight models rather than closed commercial APIs.
Built for the Daily Work of Research
Open problems in mathematics have been used as benchmarks to train the capability of many AI models. SAIR’s Open Math Model takes a fundamentally different approach. Its first phase is precisely calibrated to the real “daily pain points” of working mathematicians. The model puts practical utility first: helping researchers parse exceptionally complex, multi-page arguments, cross-reference extensive literature, explore novel mathematical examples, generate specialized code, and, crucially, advance the formalization of proofs in rigorous verification languages.
Open Math OS brings this work together in one place. Each project has its own cloud machine, running around the clock with more than 100 tools for mathematics and scientific computing already installed, Lean and Mathlib among them. When a problem demands more, Ultramath lets a researcher set several agents to work on it in parallel, while the judgment of what holds remains with the mathematician.
Democratizing the Frontier
The Open Math Model initiative is global by design. It is open to researchers across all institutions, regions, and career stages. By deliberately lowering the barrier to entry and democratizing access to advanced computing power and AI tools, SAIR Foundation ensures that brilliant minds everywhere can take part in the next frontier of mathematical discovery.
2. Safeguards for the Creators of Knowledge
SAIR’s Open Math Model is built on a rigorously defended ethical framework designed to protect the creators of knowledge.
The Community Owns Its Data
SAIR Foundation explicitly affirms that the mathematical community inherently owns its data and retains absolute authority over how it is used. That authority is exercised through the Open Math Model community, which sets the rules for how data is collected and used. The foundation operates on a strict opt-in model: only with explicit, prior consent and under formally established agreements will a researcher’s data be used for model training. Researchers hold the ultimate power to decide exactly what they share, setting firm boundaries on how AI engages with their intellectual output. Open Math OS makes these boundaries the default: projects, files, and chats remain private until researchers choose to share them, and training on their work stays off unless they explicitly consent.
Ownership and Credit Stay with Authors
Academic credit is the primary currency of the scientific world. SAIR Foundation guarantees that ownership of any independent or pre-existing work contributed to the initiative remains entirely with its original authors. This guarantee extends to Open Math OS: whatever is created there remains its authors’ own, along with the credit for its ideas. Researchers will also receive explicit, traceable attribution and recognition for the mathematical concepts, intuitions, and results they provide. The Contributor Network will record these contributions in a transparent, timestamped account of each member’s work.
Governed by the Community, Independent of Capital
The trajectory, priorities, and ethical boundaries of Open Math Model are governed directly by the mathematical community itself. SAIR Foundation openly acknowledges the necessity of partnering with industry to secure crucial computing resources, but it enforces one non-negotiable prerequisite: the complete protection of the community’s research independence. Its governance structure guarantees that the pursuit of scientific truth and academic integrity will never be subordinated to corporate profit motives or commercial timelines.
Ultimately, SAIR Foundation sets out to prove a vital premise: it is entirely possible to train world-class, state-of-the-art mathematical AI models without sacrificing researchers’ copyright or undermining the spirit of open academic reciprocity. Through this initiative, AI takes its rightful place as a true and respectful collaborator to scholars, rather than an extractive predator of their intellectual legacy.