Why Governments, Companies, and Nonprofits Must Fund Free, Open‑Source AI

Open‑Source AI Is the Next Frontier of Shared Knowledge

Takeaway: Open‑source AI should be funded by governments, corporations, and nonprofits because it preserves the collaborative knowledge ecosystem that powered the rise of modern computing and mitigates the risks of closed, proprietary models.

Open Source Powered the Computing Revolution

In the 1980s, Richard Stallman’s free‑software movement proved that sharing source code accelerates innovation.
Projects like GCC and GNU/Linux, built by thousands of contributors, became the backbone of the internet.
Transparency allowed a global community to audit, improve, and secure software, disproving the “security through obscurity” argument.

AI Is Becoming the New Closed Library

Frontier AI models are now entirely closed—the code that builds them and the training data are hidden.
Closed models turn AI into a black‑box oracle that doctors, engineers, judges, and everyday users must trust without understanding how answers are generated.
Explanations offered by proprietary models are post‑hoc stories, not verifiable audits; the same query can yield different answers over time with no way to trace the cause.

Risks of a Closed AI Ecosystem

Concentration of power in a handful of firms creates a knowledge monopoly that can steer scientific progress, policy, and public discourse.
Closed models do not guarantee safety; they can be leaked, jail‑broken, or used to amplify misinformation.
The lack of open training data means the community cannot assess bias, privacy violations, or misuse potential.

Why Openness Still Matters Even If Models Are Dangerous

"Releasing the underlying AI is not like publishing a research paper: the software itself is the capability."

The scientific community already publishes risky results (e.g., dual‑use physics) while keeping the underlying methodology open, relying on monitoring and regulation rather than secrecy.
Open AI would allow independent security audits, rapid bug fixes, and democratic oversight—benefits that closed models cannot provide.

What Kind of Openness Is Needed?

Open‑weight models – anyone can run the model.
Open‑source code and data – the full training pipeline and datasets are publicly available.

Most so‑called “open” models today only release the runtime weights, leaving the training code and data hidden. This is akin to giving users a compiled binary without the source library.

Funding Open‑Source AI: A Proven Blueprint

The open‑source software era succeeded because public compute grants, corporate philanthropy, and a default‑open rule for publicly funded research created a sustainable ecosystem.
Applying the same model to AI would involve:
- Government grants for compute resources dedicated to open‑source research.
- Corporate and nonprofit sponsorship of university labs and independent labs.
- A policy that any AI developed with public money must be released under an open license.

Community‑Driven Incentives

"We really need to band together to fund targeted inducement prizes… $200K for the first model to hit a benchmark on a fixed VRAM budget."

Prize competitions, similar to the Nobel‑style incentives suggested by Michael Kremer, can accelerate progress and provide recognition beyond monetary rewards.

Counterpoints and Rebuttals

Commercial dominance: Some commenters note that commercial AI will still dominate because developers are paid to prioritize proprietary work. Rebuttal: Open‑source projects can thrive with dedicated funding and can produce cheaper, more adaptable products, as seen in the broader software industry.
Safety concerns: Critics argue open AI could be misused. Rebuttal: Openness enables broader security scrutiny and mitigates the single‑point‑of‑failure risk inherent in closed systems.
Scale vs. utility: Frontier models are expensive, but many real‑world applications do not require the absolute cutting‑edge. Open models at modest scale can meet most needs while remaining transparent.

A Call to Action

Governments should allocate compute grants and enforce open‑by‑default policies for AI research funded with public money.
Corporations should create philanthropic programs that sponsor open‑source AI labs and prize competitions.
Nonprofits should coordinate funding, stewardship, and community governance to ensure long‑term sustainability.

By replicating the collaborative model that built the internet, we can keep AI as a public commons rather than a locked‑door library controlled by a few corporations.

David Siegel is a computer scientist, entrepreneur, and philanthropist, co‑founder of Two Sigma and founder of Open Athena. The arguments above are drawn from his July 3 2026 commentary on the need for open‑source AI.