SIIA’s Response to NTIA’s Request for Comment Regarding “Dual Use Foundation Artificial Intelligence Models with Widely Available Model Weights”

The Software & Information Industry Association (SIIA) has responded to the National Telecommunications and Information Administration’s (NTIA) request for comments on dual-use foundation models with widely available model weights. SIIA emphasizes the importance of a nuanced approach to openness and risk management in AI systems. Key points include:
  1. Defining Openness: SIIA suggests that openness should be viewed across a gradient rather than as a binary concept. They advocate for considering factors beyond just model weights, such as training data and code, in assessing openness.
  2. Understanding Dual-Use Foundation Models: SIIA discusses the definition of “dual-use foundation models” outlined in Executive Order 14110, emphasizing national security implications and the need for a balanced approach to governance.
  3. Assessing Risks and Benefits: The association examines the risks associated with widely available model weights, emphasizing the importance of considering both technical and societal impacts. They stress the need for a risk-based approach and caution against overregulation that could stifle innovation.
  4. Promoting Equity and Innovation: SIIA discusses how open foundation models can enhance equity, competition, and innovation in various sectors, including healthcare, education, and finance.
  5. Addressing Legal and Regulatory Challenges: The association highlights the importance of interoperability, standardizing licensing terms, and international cooperation in managing the risks and benefits of open foundation models.
  6. Government’s Role: SIIA recommends a collaborative approach between government, industry, and other stakeholders in setting standards, promoting innovation, and managing risks associated with open foundation models.
Overall, SIIA emphasizes the need for a balanced and flexible approach to governance that fosters innovation while addressing potential risks in the evolving landscape of AI technology.

Comments are closed.