Blog/Company
Latitude.sh has signed the Open Weights and American AI Leadership letter
Latitude.sh has joined more than 230 companies, researchers, and industry organizations in signing the Open Weights and American AI Leadership letter.
The letter makes a simple argument: AI leadership won't be decided by a single frontier model. It will be decided by whether developers, businesses, and public institutions can build with AI on their own terms. It calls for policies that keep open-weight models viable, competition healthy, and investment in AI infrastructure flowing.
Why open weights matter
The comparison the letter draws is to open source software. Open source gave everyone a shared foundation to build on, and most of the internet now runs on the result. It proved that openness compounds.
Open weights offer the same thing for AI models. Download it, inspect it, adapt it, run it on infrastructure you control. No training run from scratch. No single provider deciding what you get.
That matters more as AI moves out of experimentation and into production. Not every workload needs the largest frontier model. Plenty of workloads run better, and cheaper, on something smaller and specialized.
Matching the model to the task is what makes AI economics work at scale, which brings us to the letter's sharper point: leadership won't be measured by who ships the biggest model. It will be measured by how far AI can effectively reach— into healthcare, education, manufacturing, financial services, and government.
The real opportunity for growth is in ordinary problems that can be solved at scale.
The power of choice
Open weights create competition among model developers, and downstream among cloud providers, infrastructure platforms, and everyone building on top. Competition is what gives customers control.
Organizations have different technical, operational, and regulatory constraints, and no single deployment model covers all of them. As AI matures, most teams will mix: multiple models, multiple providers, dedicated compute, and storage chosen for the workload. Infrastructure should make that easier rather than narrower.
Openness comes with responsibility
The letter doesn't pretend openness is free of risk. Once weights are released, models can be modified and redistributed, raising real security and governance questions.
But its answer is the open source answer: collaboration, transparency, evaluation, and responsible innovation rather than restricted access.
Open source got more secure because a global community kept reviewing it. Open weights let researchers do the same to AI models: probe them, find the failure modes, strengthen the safeguards.
Why we signed
One of the things that resonates most with me about this letter is the recognition that AI leadership isn't just about building larger models. It's about creating an ecosystem where organizations have genuine choice in how they develop and deploy AI.
As AI adoption accelerates, customers need flexibility, not only in the models they choose, but in where those workloads run and how they connect compute, networking, and storage. We've always believed infrastructure should adapt to customers, not the other way around, and supporting the principles outlined in this letter reflects that commitment.
Success in AI won't come down to model development alone. It will come down to whether organizations can deploy those models efficiently, securely, and at scale, wherever it makes sense to run them.
And that's the part we work on.