The White House is reportedly preparing to incorporate open artificial intelligence models into an updated policy framework, according to sources familiar with the discussions. This move would represent an expansion of the administration's approach to AI regulation, which has previously focused more heavily on proprietary, closed-source systems. The potential inclusion of open models reflects an ongoing effort by the White House to establish comprehensive guardrails for AI development and deployment.
The discussion around open models follows earlier actions by the administration to address AI risks. In October 2023, the Biden administration issued Executive Order 14110, which outlined policy goals for safe, secure, and trustworthy AI development and use. This order, considered a significant piece of AI governance, directed federal agencies to take various actions, including creating chief artificial intelligence officer positions. However, this executive order was rescinded by President Donald Trump in January 2025.
More recently, the Trump administration signed an executive order in June, directing agencies to create a benchmarking process to assess the advanced cyber capabilities of AI models within 60 days. This order also established a voluntary pre-deployment review regime, asking developers of "covered frontier models" to grant the government up to 30 days of access to their most advanced systems before public release. This access is specifically for evaluating the models' potential in cyberattacks. By early August, administration officials indicated that this framework was complete and on schedule.
However, the current framework for cybersecurity testing has reportedly exempted open-source and open-weight AI models from federal review, concentrating instead on advanced closed-source systems. This voluntary review process is said to apply only to "covered frontier models", closed-source AI systems with state-of-the-art capabilities that could pose national security or cybersecurity risks. Companies like OpenAI and Anthropic are expected to fall under this framework. The White House has not publicly released the full framework, with details reportedly shared only with the companies involved.
The National Telecommunications and Information Administration (NTIA), a bureau within the Department of Commerce, has also been active in developing AI policy. In March 2024, the NTIA released its AI Accountability Policy Report, which called for independent audits of high-risk AI systems and recommended federal guardrails, including additional disclosures. The report suggested the creation of "AI disclosure cards" similar to nutrition labels, detailing aspects such as the product's name, model type, data retention frequency, and whether a human is involved in its operation. The NTIA's recommendations also included strengthening the government's capacity to address AI-related risks across economic sectors and maintaining a registry of high-risk AI deployments, adverse incidents, and system audits. This report followed a request for comment issued by the NTIA in April 2023, which sought feedback on policies to support AI audits, assessments, and certifications.
The debate surrounding open versus closed AI models has divided the technology industry. Some companies emphasize the potential dangers of widely accessible AI model components, while others highlight the importance of open science for researchers and startups. Companies such as Meta Platforms and IBM have been vocal proponents of an open approach. The White House's decision to exempt open models from its initial cybersecurity review framework was seen by some as an incentive for U.S. firms to invest in competing open models, especially as Chinese firms have released open models comparable in capability to leading American commercial AI technology.
The potential expansion of the White House's AI policy to include open models suggests a recognition of their growing influence and the need for a more comprehensive regulatory scope. This shift could lead to new guidelines or requirements for developers of open-source AI, potentially introducing new considerations for innovation and security in the sector.