AI sovereignty means owning the models, the tools, and the infrastructure your organization runs its intelligence on, instead of renting them from a vendor who can see, shape, or dispute your work.
The OpenAI-Navier Stokes controversy just showed what happens without it: a mathematician’s private work and a competitor’s black box collided, and nobody outside either building could say for certain what happened.
That’s not a hypothetical risk anymore. It’s this week’s news.
What Did OpenAI and the NYU Mathematician Actually Fight About?
NYU mathematics professor Tristan Buckmaster announced proofs toward one of math’s seven Millennium Prize Problems, working with Anthropic mathematician Levent Alpöge and using AI models including OpenAI’s Codex.
Buckmaster then says a parallel OpenAI effort, using a similar approach almost nobody else in the field was pursuing, caught up before his own results were public.
Buckmaster’s own statement, reported by TechCrunch, describes an OpenAI mathematician pressing him to drop Alpöge’s credit because Alpöge works at a rival lab, and warning him, “Why would you ruin your career?” OpenAI disputes that characterization and says it never saw the pair’s unpublished work through any channel.
Days later, OpenAI announced its own full proof, produced by an unreleased model over roughly a week of computation. Nobody outside OpenAI has independently reviewed it yet, and neither side’s account of the underlying data question can be fully settled from the outside. That’s the part worth sitting with.
Why Does This Matter to You?
Strip out the personalities and the story is a sovereignty problem. Buckmaster ran his most sensitive, career-defining work through infrastructure that he didn’t own, using a model he didn’t control, and then had no way to independently prove what did or didn’t happen to it on the other side.
That’s true of every organization sending proprietary documents, contracts, or data through a third party’s AI API. You can trust the vendor’s word, or you can own the infrastructure and never have to ask.
Why Should an Organization Get off Per-Token Economics?
Every prompt you send to a hosted AI model gets broken into tokens and billed by the piece, on hardware you’ll never see, under terms the vendor can change. OpenAI’s own account of its Navier-Stokes effort put a number on what that looks like at scale:
- About 300 billion output tokens in one week
- Roughly $22.5 million in compute at current rates
That’s not a criticism of the math. It’s a preview of the bill.
A metered, per-token relationship with your AI provider means you never actually own anything. You’re renting intelligence at a rate somebody else sets, indefinitely, and the meter never gets cheaper on its own.
Owning your inference layer flips that math. It’s a fixed cost you control instead of a variable one you don’t, and it’s the difference between building an asset and renting one forever.
Why Does Using Your Own Tools, Models, and Solutions Matter?
BIS has been self-funded since 1986. No investors, no debt, no board seats sold, ever. That independence is exactly why we can say this plainly: the ability to run your own AI stack, on your own terms, is worth more than whatever convenience you give up by not using someone else’s.
We’re not anti-cloud AI. Grooper works with leading cloud LLMs, including OpenAI, as one deployment option among several. But it also runs on models hosted on your own GPUs, or on ours, or on no model at all when a deterministic rule beats one.
That optionality is the point. Sovereignty means you get to choose, not that you’re locked into a single answer forever.
How We’re Practicing What We Preach
We don’t just sell the idea. We run it. Here’s how:
We use our own AI to build our own software. Aristotle, our internal AI server, writes code for Grooper today.
We’re not describing a hypothetical internal tool. We’re running one, on infrastructure we own, developing the platform we sell. Our Aristotle dashboard tracks all usage and shows us how much we have saved.
Grooper is a Model Context Protocol (MCP) hub. Grooper exposes its own capabilities as MCP tools across categories like batch management, search, event logging, and system administration, so AI agents, whether Claude, ChatGPT, or an assistant you built yourself, can operate Grooper directly. That’s sovereignty applied to the interface layer: you choose which agent talks to your data, instead of being locked into one vendor’s proprietary agent framework.
We move our own data in and out of Microsoft Dataverse using our own tools, not whatever import and export utility a platform vendor decided to ship. Owning that pipeline means your data stays portable on your terms, not a vendor’s release schedule.
What Should Your Organization Do Next?
Ask who actually controls each layer of your AI stack:
- The data
- The model
- The infrastructure
- And the tools that move information between systems
If the honest answer is “a vendor we’ve never audited,” that’s a sovereignty gap, not a compliance detail.
You don’t have to build everything yourself to close it. You have to know what you’re renting, what it costs over time, and what happens if the terms change or the dispute goes public.
BIS has spent decades building the platform, and the practice, to help organizations own that answer instead of hoping it never comes up.
You don’t have to sort this out alone, and you don’t have to guess at what “sovereign” should mean for your organization. BIS has spent decades running the hardest document and data problems other companies wouldn’t touch, then built Grooper to run them ourselves, on our own AI, before we ever offered it to a client.
Let’s talk through what a sovereign AI setup would actually look like for your data, your documents, and your infrastructure. We’ll tell you plainly what that takes, what it costs, and what it doesn’t require you to give up.
