OTI Urges NTIA to Encourage the Development of a Strong Open-Source AI Ecosystem in the United States

The Open Technology Institute at New America submitted comments yesterday in response to the National Telecommunications and Information Administration (NTIA)’s request for public submissions on the risks, benefits, and policy and regulatory approaches relating to “Dual-Use Foundation Artificial Intelligence Models with Widely Available Weights,” emphasizing the importance of policy and regulatory approaches that encourage the development of a strong open-source AI ecosystem in the United States.

In our comments, we offer four principles and recommendations for U.S. government and related stakeholders developing policy and regulatory approaches to open models:

1. Study marginal risk and articulate harms with specificity.
2. Create common policy and regulatory requirements that apply to foundation models no matter where they fall on the spectrum of openness.
3. Consider the broad range of relevant national security and foreign policy objectives before recommending sweeping policy or regulatory action aimed either at vaguely defined or very specific defined security risks.
4. Develop a thoughtful approach to cybersecurity software liability that accounts for the need to incentivize innovation in open AI models.