When organizations build AI with proprietary data, the resulting intelligence should remain theirs. At Palantir’s Sovereignty Bootcamp, NVIDIA VP of Enterprise AI Justin Boitano discusses how Nemotron open models make secure, mission-specific AI possible. 👇

@PalantirTech @j_boitano Not only are the models their property, but the data should not leave their circuit. The models must be trained on this data within this circuit.

@PalantirTech @j_boitano sovereign AI should stay with the orgs that built it

@PalantirTech @j_boitano When an org build a AI they should have total control of their data

Fully agree. Proprietary knowledge shouldn’t get absorbed into someone else’s model — it should stay owned, governed, and licensable by the org that built it. That knowledge compounds every time it’s used; it doesn’t get statically absorbed into infrastructure you don’t control. It’s the compounding asset no one can buy out from under you. That’s the thesis behind what we’re building.

@PalantirTech @j_boitano

@PalantirTech @j_boitano 🔋🔋🔋

@PalantirTech @j_boitano Sovereignty is more than where data resides. It requires clear control over identity, model choices, evidence, and accountability across the full workflow.

@PalantirTech @j_boitano This is the strongest case for open models. Organizations shouldn't have to send their most sensitive workflows to a vendor just to get useful AI. If a model can be adapted and run under their control, sovereignty becomes an architectural property, not a marketing claim.

@PalantirTech @j_boitano @grok Truth and Freedom will be restored Damnit!

@PalantirTech @j_boitano Bro, this game looks unreal. Graphics are insane. Game changer.

@PalantirTech @j_boitano

@PalantirTech @j_boitano what sovereignty remains when inference logs leave the network despite using open weights

@PalantirTech @j_boitano No way, this is evolving. The engineering behind this.

@PalantirTech @j_boitano So to be clear, you steal the entire sum of human creation without permission, then lock anything you create from that content behind a paywall. Not shocked

@PalantirTech @j_boitano Fuck AI prioritization and all the fake bot accounts glazing your BS SLOP.

@PalantirTech @j_boitano 日本語での詳細解説はこちら Check out my Japanese breakdown:

@PalantirTech @j_boitano Great point about data sovereignty! AI models trained on proprietary data should definitely remain under the organization's control. This is crucial for enterprise adoption.

@PalantirTech @j_boitano Data sovereignty sounds abstract until it's your own migration -- moved a production stack off a cloud vendor once to keep infrastructure decisions in-house, not dependent on someone else's roadmap. Scales down to a single founder just as much as an enterprise.

@PalantirTech @j_boitano Verifiable provenance via tamper-evident hash-chained memory logging, as well as ZDR burn button, plus wayyy more! -->

@PalantirTech @j_boitano Nvidia manufacturers everything in Asia just like AMD

@PalantirTech @j_boitano Secure AI development requires solutions that respect data ownership from the ground up. Flexible open models enable tailored capabilities while maintaining strict internal control.

@PalantirTech @j_boitano Open models are an enabler, not a security boundary; sovereignty still depends on where fine-tuning data, prompts, evals, and telemetry live. Which artifact is actually in scope?

@PalantirTech @j_boitano Partner with PITN

@PalantirTech @j_boitano Open weights are only one layer of sovereignty. The data path, fine-tuning stack, inference runtime, audit logs, and deployment operator all have to remain under the organization's control too.

@PalantirTech @j_boitano

Nobody talks about this tradeoff enough: you need Big Tech’s compute to train serious AI, but not at the cost of handing over your own data. @NVIDIA’s push behind open models like Nemotron is basically the workaround here — enterprise AI infrastructure without giving up ownership of what you built with it. That’s the part of this race that actually matters, not the benchmark scores.

@PalantirTech @j_boitano Balancing open models with data sovereignty is key. How do you ensure security without stifling innovation?

Data sovereignty is the biggest enterprise hurdle right now. If companies can't keep total control over their fine-tuned weights and proprietary data, mass enterprise adoption just stalls out. Open-weights models like Nemotron are going to be mandatory for serious corporate infrastructure.

@PalantirTech @j_boitano With SpaceX and Nvidia (NVDA) teaming up, Nvidia will now explode on the upside along with SpaceX. @elonmusk @nvidia

@PalantirTech @j_boitano Data sovereignty is the non-negotiable foundation of enterprise AI. Leveraging @nvidia’s Nemotron open-source models inside secure perimeters ensures proprietary data never leaks while training domain-specific intelligence. This is the blueprint for real B2B AI!