National AI Action plan should expand open-source offerings, respondents say
Part of President Donald Trump’s executive order on AI, the RFI asked for input on elements that the forthcoming AI Action Plan should include, particularly the ways it which it can help remove barriers to American leadership in AI technology development and deployment.
A number of influential entities across the private, advocacy and academic sectors said that enabling open-source software regimes is key to that mission.
Open-source software refers to foundational code that is made publicly available to developers. This allows original source code to be manipulated and used for other software tools and offers greater transparency into the software’s inputs — such as training data, parameters and model weights — and how they work together.
As more advanced, generative AI systems continue to evolve, open-source software enables both the democratization of knowledge regarding how the software is developed, and where it may be vulnerable. Expanded access to advanced models also offers under-resourced developers and entities the chance to gain experience and build tools with AI.
Advocacy organizations like the Open Source Initiative offered a strong rationale for building a strong open-source ecosystem in the U.S.
“At a time when AI is evolving rapidly, the U.S. government’s AI Action Plan should include Open Source AI as a foundation to ensure as many minds as possible — from startups and researchers to major tech firms — are working to find the next Innovation,” OSI’s comments read. “Driving collaboration through Open Source can bolster the competitiveness of US companies on the world stage.”
Other advocacy organizations and academic entities agreed. The Federation of American Scientists submitted comments saying that the White House should use its power to bring leading AI companies, academic institutions and government agencies together to jointly develop open-source tools.
“These partnerships could encompass co-funded research, shared datasets and computing resources, collaborative access to advanced AI models, and joint development of open-source tools,” the FAS wrote. It added that continued public-private partnerships will help align national priorities and expedite specific solutions to these issues.
On the academic side, the MIT Computer Science and Artificial Intelligence Lab added that enabling open-source licensing regimes has helped technology innovate at a rapid pace for decades.
“Today, we should maintain the healthy mix of open source and proprietary licensing modes for AI models, data and underlying software,” MIT CSAIL wrote. Their comments also addressed concerns over security and bad actors exploiting vulnerabilities in open-source code, citing research that developed new styles of limitations for how models can be applied.
“Still, neither open nor closed-source licensing will prevent software-based AI algorithms or other expertise from falling into the hands of determined, well-resourced adversaries,” CSAIL said.
Leading private sector companies also voiced support for expanded open-source licensing options. IBM notably said that U.S. acceleration of open-source AI will create more benefits than security risks.
“Open models encourage greater involvement and scrutiny from a larger community of stakeholders, increasing the likelihood that bias and vulnerabilities are identified and patched,” IBM wrote in its comments. “Ultimately, when a broad community is represented in AI development, safety is prioritized. These benefits are not confined to the commercial sector, but extend to the national security and intelligence community as well.”
IBM’s conclusion recommended that the Trump administration enable open innovation in AI by offering a repository of computing and data resources for smaller entities to participate in the sector, potentially through funding federal programs like the National AI Research Resource.
OpenAI, Mistral AI, and MITRE echoed the need for more open-source AI software resources. OpenAI added that the strategy should help expand the open-source model with a strong export control regime to ally countries that commit to U.S.-led democratic values to guide responsible AI development.
“We believe the question of whether AI should be open or closed source is a false choice—we need both, and they can work in a complementary way that encourages the building of AI on American rails,” OpenAI wrote. “Making sure that open-sourced models are readily available to developers in these countries also will strengthen our advantage.”
This is not the first time the subject of open-source AI development has been presented to the White House. In 2024, the National Telecommunications and Information Administration issued a landmark report telling then-President Joe Biden that all-purpose foundation models should publicize the weights they use to calculate their outputs.
“NTIA’s report recognizes the importance of open AI systems and calls for more active monitoring of risks from the wide availability of model weights for the largest AI models,” lan Davidson, the former assistant secretary of Commerce said last year in a press release. “Government has a key role to play in supporting AI development while building capacity to understand and address new risks.”
]]>