Recent policy shifts by the Trump administration imposing new restrictions on the release of private AI models are intensifying the call for open-source alternatives.
President Trump’s administration has implemented measures to curb the dissemination of AI models developed by companies like Anthropic and OpenAI. This action includes a regulatory “kill switch” for models under exclusive control of a single organization and built on proprietary data.
Proponents of open-source AI argue that the government’s unprecedented regulations on cutting-edge AI labs might inadvertently benefit China. China is known for providing less expensive, open-source AI models that are available to individuals and businesses globally.
AI specialists are asserting that this scenario shines a light on the importance of bolstering open-source AI development within the United States. Expanding these efforts is seen as crucial to ensuring China doesn’t gain an upper hand due to a dearth of pioneering American AI models.
“Most businesses already function within environments that utilize multiple models and agents. They require a control plane they can manage—and sometimes even own,” commented Felix Van de Maele, CEO of Collibra, a company specializing in data intelligence.
“When a company gets 90 minutes to pull a model deployed to hundreds of millions of people because of a competitor’s complaint, that need becomes urgent fast,” Maele said. “This didn’t create the anxiety we are all experiencing now. It removed any remaining doubt.”
Maele was referring to how the Trump administration gave Anthropic 90 minutes last month to take down its latest Claude Mythos 5 and Fable 5 models after Amazon raised cybersecurity concerns. Anthropic complied with the export control order, and the two models went dark for more than two weeks until the administration lifted the restrictions last week.
The Trump administration later asked OpenAI to delay the public rollout of its new GPT-5.6 series amid similar concerns, though it was not an export control order.
The situation exemplified how private AI firms could be pressed to cut off thousands, if not millions of customers from their models with the flip of a switch.
Unlike private models, open-source models live in the public domain where any individual or business can download and customize them for personal use. These systems can be used, modified, examined and shared with anyone, for any purpose.
Proponents often consider open models to be more transparent, as the entirety of training data, code and process is publicly available.
In some instances, a model may not be open-source, but can have open weights, meaning the ways a model is trained to sift through information and formulate answers are made public.
“You are able to see the weights, you are able to see what is running, how these models are operating,” another AI industry executive told The Hill. The models “allow for deep auditing and safety testing and bias mitigation, whereas closed models are black boxes that are controlled by a private company.”
The Trump administration has increased its oversight over private AI models, in a sharp shift from its light-touch regulatory approach last year.
Trump signed an executive order last month laying out a voluntary process for government testing of models before release. It also called on federal officials to establish safety standards for new AI models meant to supplant state laws and give firms clear guidelines.
While he emphasized no government testing would be mandatory, the pressure put on Anthropic and OpenAI suggested otherwise. Open models can be tested at any point by the government, without a formal process.
OpenAI has two open-weight models called gpt-oss, though they are considered much less powerful or advanced than its private ones.
Anthropic and its CEO Dario Amodei have long opposed open-weight models, warning during a 2023 Capitol Hill visit that the scaling of open source models is “going down a dangerous path.” He argued private models allow changes after they are deployed, including the revoking of a user’s access.
“When you control a model and you’re deploying it, you have the ability to moderate usage. It might be misused at one point, but then you can alter the model,” he said at the time.
While private models like ChatGPT, Claude and Gemini are surging in popularity among the public, they are also the priciest and unattainable for companies that do not want to spend tens of thousands of dollars.
The technology and researcher community have long pushed for open-source alongside private models, but experts said the latest events crystallized the need for more open models.
Following the export control order against Fable, Box CEO Aaron Levie said the situation is getting “to the core of one of the central debates” in AI.
“If open weights AI can remain a close second to frontier intelligence, then the equation reverses. With a highly regulated approach, you’ll own the frontier market still, but the vast majority of tokens used will go to an alternative stack,” Levie wrote on X.
“That stack will include the model and the underlying hardware that runs it, in the limit. And that stack will be controlled and monetized by someone else,” he added.
A token is a unit of measurement for AI usage. Costs for running frontier models, and using more tokens in turn, are rising, prompting some enterprises to shift towards cheaper, open-source models.
The so-called “tokenmaxxing,” practice, where users maximize token usage as a way to track productivity, is contributing to the rising costs of these frontier models.
A study from MIT Management earlier this year found open-source models accounted for a fifth of all AI token usage.
Alex Karp, the CEO of technology giant Palantir, recently slammed the token structure being used by frontier labs like OpenAI and Anthropic, pushing instead for open models.
“I’m not throwing shade at them, but something has gone completely wrong and the basic view among enterprises in this country is I’m going to ‘chillax’ and waste my time with tokens, I’m going to get no value and they’re going to take my IP,” Karp said on CNBC’s “Squawk Box” last Wednesday.
“What the technical customers want…control over their compute, their models, their data stack and their alpha. They want to know they own the means of production, it’s not being transferred to someone else,” Karp added.
The difference in capabilities between frontier and private models are widely debated. While frontier models may have more advanced use cases, companies looking for specific uses rather than the whole package will turn to cheaper models, experts said.
Vinny Troia, the founder and CEO of intelligence firm Shadow Nexus, said this “price gouging” is more of an incentive to look at open models, telling The Hill, “What’s happening with the White House is only going to further compound the problem.”
In the interim, Chinese open source shows no signs of stopping.
According to research from venture capital firm Andreessen Horowitz, 80 percent of developers around the world who use open-source tools are building with Chinese models.
These companies include U.S.-based firms. Airbnb CEO Brian Chesky, who said last year that the company is “relying a lot” on Alibaba’s Qwen model, describing it as “very good” and “fast and cheap.”
“We’re starting to see the open models coming out of China…catch up with the frontier, and for a while there’s been this narrative of America winning the AI race. While [the U.S.] has been winning on capability with the best models, we’ve been losing,” said the AI industry executive.
