Science drives many of the things you take for granted: cleaner energy, better medicine, advances in manufacturing and national security.
What if we could speed up decades of discovery into just a few years? OpenAI proposes bringing frontier models together with labs, supercomputers and scientists to try exactly that.
What OpenAI proposes and who it's working with
OpenAI announced concrete support for the Department of Energy's Genesis mission, working with 17 National Laboratories, universities and research teams. The idea isn't to produce isolated models, but to integrate AI into the tools and workflows scientists actually use.
The key commitments include:
- $4 million in access to Codex for roughly 2,000 researchers in labs and universities, to improve programming and computational reasoning.
- $3 million in API support for two large-scale scientific campaigns that will test frontier models against major challenges.
- An offer of up to $10 million in API usage for every $2.5 million spent by participating researchers, to scale practical capacity.
- Access to GPT-Rosalind for selected researchers on eligible bioscience projects.
- Early access to models and features for lab leaders, and expanded access to advanced cybersecurity capabilities for trusted researchers.
How it would work in practice
This isn't just about handing over models: it's about walking alongside scientists in their daily work. That means integrating AI with federal data, simulations, supercomputers and experimental equipment.
A concrete example: at Los Alamos an advanced reasoning model was deployed on Venado, their supercomputer, as a shared resource for nuclear security labs. That shows how AI can collaborate with high-performance computing, not replace it.
Another practical piece was the "AI Jam Session" with nine National Laboratories and more than 1,000 scientists, where they tested models on specific problems and gave structured feedback. It's a step toward understanding what works and what still needs human oversight or more research.
Initial scientific campaigns
OpenAI will support two broad initiatives from the start:
- A campaign to advance high-temperature superconductors, combining AI, simulation, materials and experiments. Imagine materials that work at higher, more practical temperatures and pressures: impact on energy, transport and medicine.
- An "Atlas of the Machine-Accessible Frontier" to map where AI can already drive scientific progress with existing data and knowledge. This helps distinguish problems AI can tackle today from those that still need experimental work.
These campaigns aim to develop repeatable methods for applying AI to big national challenges, not one-off solutions.
Why this matters (and what limits it has)
The bet is strategic: national advantage won't come from a single model, but from sustained ability to apply advanced AI to real problems. That requires coordination: government funding and priorities, labs defining problems and ensuring rigor, and industry providing capabilities and deployment.
Does this mean AI will solve everything by itself? No. Researchers remain central: they ask the questions, validate results and check for unintended effects. Projects in bioscience and security also demand rigorous risk assessments and human control.
Risks, governance, and trust
OpenAI acknowledges risks and is already working on safety evaluations in real lab environments. In bioscience, for example, joint tests were developed with Los Alamos to understand both opportunities and dangers of multimodal systems.
Collaboration with public institutions also seeks to open reliable access channels: early access to models for lab leaders, safety controls and support to integrate AI without endangering critical facilities.
Why should this matter to you, even if you're not a scientist?
Because the relationship between AI and science shapes technologies that become the products and services you use: better medical diagnostics, more efficient materials, and more resilient power grids. If these capabilities are integrated responsibly, the economic and social impact can be huge.
And if you're curious or have projects at a university or lab, this opens concrete paths to test and scale ideas with technical and financial support.
OpenAI commits to working with the scientific community and public institutions so frontier AI stops being just a promise and becomes reliable, repeatable strategic infrastructure. The goal is to accelerate discoveries, but with scientists in charge and clear safety criteria.
Original source
https://openai.com/index/advancing-the-next-era-of-national-science