The AI race will not be won by whoever locks the best model in a vault. It will be won by whoever puts capable models in the most hands.
On July 24, 2026, an unlikely coalition—chipmakers, cloud giants, defense contractors, venture firms, and open-source foundations—publicly agreed on one thing: freely available model weights are essential to American AI leadership. Companies that compete fiercely with each other signed the same statement. Why? Because the deeper logic of open ecosystems, proven over thirty years of open-source software, is now playing out in artificial intelligence—and the stakes are national.
Open Weights turns that one-page industry statement into a full strategic argument. Shane Larson walks through what open weights actually are (and are not), why the economics of AI diffusion favor openness, how concentration in a handful of closed labs creates fragility rather than strength, and why the loudest safety arguments for restriction often get the security question exactly backwards. He takes the distillation debate seriously, examines China's aggressive open-weight strategy without hype or panic, and lays out concrete policy prescriptions: compute access, shared public assets, and the discipline to avoid premature restrictions that would cede the global open ecosystem to someone else.
This is not cheerleading for any company or any model. It is a clear-eyed case that startups, universities, hospitals, factories, and governments need the ability to download, inspect, adapt, and self-host advanced AI—and that the country whose ecosystem makes that easiest will set the standards everyone else builds on.
What you'll learn:
- What open weights actually are—and how they differ from open source, open data, and marketing labels
- The open-source software parallel: how Linux, and the stacks built on it, predict where AI value will accrue
- The economic case: why diffusion into the real economy beats frontier benchmark wins
- How concentration risk threatens competition, resilience, and national security at the same time
- The security argument for transparency—and why obscurity fails as a defense strategy
- The distillation debate, China's open-weight push, and what a serious U.S. response looks like
- Practical implications for builders, enterprises, investors, and policymakers
This book is for you if:
- You shape, advise on, or advocate for AI policy and need the full argument, not a slogan
- You lead technology strategy and must decide where open models fit in your stack
- You build or fund AI companies and want to understand where value will actually be captured
- You read serious tech-policy nonfiction on semiconductors, export controls, and open source
The most consequential AI decision America makes may not be about any single model—it will be about who gets to build. Read the argument before the rules get written.
