“What you can never allow to happen is you pay somebody and then as a reward for you paying them, they get to replicate your business.”
Palantir CEO Alex Karp Opening Remarks at Sovereignty Bootcamp Jul 27, 2026, at 3:25 speaking about Closed proprietary AI models
“Open weights let every organization match the right model to the right job at the right cost…”
“Open weights let every organization match the right model to the right job at the right cost, reserving frontier-scale capability for genuine frontier problems and running efficient, specialized models everywhere else. That discipline is what will make AI economically sustainable as its use scales into billions of everyday tasks. America wins the AI era by diffusing it into the workflows of factories, hospitals, farms, classroom, and main street businesses.” Jensen Huang, July 24, 2026 on x: Open Weights and American AI Leadership, paragraph three.
Token Costs and Protection of Trade Craft – IP
This rise is being driven by factors like token costs and the desire/concern for protecting trade craft and IP from unintended use or exposure.
There are two sides to this, The contractual side and the AI-model-hardware side.
The Contractual Side: Closed models (Claude, ChatGPT, Copilot etc.) are emerging as a potential threat to the security of your knowledge and tradecraft embodied in data exposed to and generated from your use of their AI model. Using Claude, for instance even with a Zero Data Retention (ZDR) agreement in place, has holes in the protection of your secret sauce recipe. Cowork and third-party API integrations for example, are two of several features that are not eligible for ZDR.
There are myriad agreements and sub-agreements involved to truly understand the hosted model provider’s and sub-processor’s obligations. A daunting amount of legal and other work is needed to secure sufficient controls. These controls apply to the data exposed to and generated from your use of their AI model.
The AI-model-hardware side focuses on a protective AI-use formula that safeguards your trade-craft and other vital information. It relies on open-weight AI models running on hardware you control. Options exist for this approach. EUS is committed to leading in this emerging wave of open-weight models. These models run on hardware our clients control.
We partner with frontier AI hardware and services providers to craft both the hardware AI infrastructure and data flow, as well as guide the soft AI infrastructure, the people using it. We do this by careful examination of work flows and data involved as well as contractual assessments of any closed model providers in use.