National Defense

There’s A New Development Blowing Up America’s AI Debate

(Photo: NASA/GRC)

A rising open-source American artificial intelligence (AI) industry seeks to challenge the prevailing narrative that companies and consumers must choose between closed, frontier labs and Chinese alternatives.

Politicians and experts have long focused on the competing spheres of AI development: open-source Chinese and closed, frontier AI labs such as Anthropic and OpenAI. Now, many leading tech companies are launching open-source AI models that give consumers and corporations the benefits of open AI — customization, lower cost and privacy — without relying on Chinese AI.

“We are dying for an American alternative,” Michael Frank, the CEO of Radiant Intel, an AI platform that does geopolitical risk analysis and a visiting AI fellow for the Wilson Center, told the Daily Caller News Foundation in an interview about the need for American open-source AI.

Open-source models are AI systems that can be altered and customized without having to request permission from the AI lab and are often free to use, according to IBM.

“The vast majority of AI deployments are going to be built on open source AI models, and the reason for that is because we’re headed towards the agentic era,” Frank continued, referring to AI “agents” that can perform tasks for the AI user, per IBM.

The growth of the American open-source AI industry may undermine calls for regulation that may ultimately protect the leading, closed-source frontier AI labs such as Anthropic and OpenAI, the Radiant Intel chief executive said.

“That’s the problem with the framing: with China being seen as open source and American frontier AI being closed source, we feel like we have this duty to protect this nest egg, but we shouldn’t do that because that’s not where we’re headed in the long run, and it is not the government’s responsibility to give these guys a business model. We should let the market decide,” Frank explained.

“The reason why open source is important is because you need to be able to customize and configure. It’s just a pure economic calculus. Why are we going to pay a 3x markup for a model?” Frank said, referring to the higher cost of closed, frontier AI models.

Other experts believe American open AI may help businesses with lower cost solutions.

“A strong American open-source ecosystem finally gives businesses a real alternative. Open models cost a fraction of commercial models, but do not need to outperform every frontier model on every task,” Lauren Gil, the governing board member of the Tokenomics Foundation as well as the cofounder and president of Kimchi Coding, told the DCNF.

“You would not rent a Ferrari to bring home groceries. Likewise, a routine coding task does not automatically require the most expensive model,” Gil remarked.

“We recently launched Kimchi Coding to tackle this specific problem head-on. It automatically matches coding tasks to the right models, so developers can spend more time building without worrying about each request’s cost,” Gil added.

Gil added that open-source models allow corporations to better secure their most sensitive data than closed models.

“It is all about trust and security. Commercial models have a trust issue,” he explained. “We all know closed commercial-model companies use any training data they can get for their next release. As a result, some enterprises are reluctant to share their most sensitive intellectual property, data, and source code with service providers that may use that data for training, willingly or unwillingly. An American open-weight model gives enterprises the comfort of running it on-prem, air-gapped, on their own dedicated GPUs, or within a secure cloud environment they control.”

Open-source AI was once dominated by Chinese labs and their models experienced so much demand that the Trump administration considered banning or limiting access to them over the summer, the DCNF reported.

In July Moonshot AI, a Beijing-based startup, launched Kimi K3, an open-source model that threatened America’s lead in the global race on AI, according to Axios.

The Chinese open-source race became predominantly represented by Chinese labs such as Moonshot, DeepSeek, and Z.ai, according to Kyle Chan, a fellow at the Brookings Institution. The advent of Kimi K3 made the Trump administration appear more interested in banning Chinese open-source AI over the summer, according to Axios.

Now, Reflection, an Nvidia-backed startup developing its own open-source AI model, enters the national dialogue.

Reflection’s open-source model, Beam, could rival the most expensive closed, frontier AI models from Anthropic and OpenAI and keep pace with Chinese open-source models, Axios reported this week.

Reflection CEO Misha Laskin has argued that owning and controlling how the AI operates may be more important than raw AI power.

“Once you’re spending that amount on intelligence, you want to move from renting it to owning it yourself. That’s kind of where open source is very powerful because it is customizable at every level,” Laskin told Axios in October.

“Our goal is to give end users control, sovereignty and continuity, at lower cost – and contribute to a growing open ecosystem, with multiple frontier labs pushing one another forward. The launch of Beam is just the start,” Eleanor Hawkins, a public policy communications official for Reflection, wrote in October on X.

Introducing Beam: @reflection_ai ‘s first open-weight model.

Our goal is to give end users control, sovereignty and continuity, at lower cost – and contribute to a growing open ecosystem, with multiple frontier labs pushing one another forward. The launch of Beam is just the start.
https://t.co/WrlmMUbEWb

— Eleanor Hawkins (@emayhawk) October 5, 2026

Nvidia CEO Jensen Huang has long sought to combine high-end computing power with open-source AI models so that businesses could retain their own data, according to The Information.

An Nvidia spokesperson referred the DCNF to Huang’s comments in late September.

“I believe very deeply in advancing open intelligence,” he told CNBC at the time. “I think we ought to democratize intelligence. I think every company will become an AI company. And this platform really, really keeps that community thriving.”

There are other rising open-source AI models.

Nvidia in August announced a $1 billion investment in the AI startup Poolside to build its own open-source models that would compete against Anthropic and OpenAI, according to the Wall Street Journal. Poolside employees will help Nvidia develop the Nemotron project, which aims to develop its own open-weight models that the company released in 2023.

Thinking Machines Lab in July released its open-source model, Inkling, according to Forbes. The AI startup said, although its model is not the strongest, it is very customizable, according to a Thinking Machines Lab press release.

Thinking Machines Lab declined to respond to a DCNF comment request.

Together AI runs a cloud-based platform for running, fine-tuning and scaling open-weight AI models, per a blog post on its website. The AI startup did not respond to a comment request from the DCNF.

The Trump administration appears to be on board with open AI as well.

A Reflection spokesperson referred the DCNF to Laskin’s panel with White House National Cyber Director Sean Cairncross at the Black Hat conference in August.

“I think the role that open source plays is vital to this ecosystem,” Cairncross said at the Black Hat during the panel about how open-source AI plays a part in cybersecurity.

“We are extremely interested in looking at ways to build U.S. open source, making it competitive, making it the preferential adoption by planet Earth for U.S. open source,” he continued.