![NTIA Seeks Feedback on “Risks and Benefits” of Open-Weight AI Models](http://media.assettype.com/cdomagazine%2F2024-01%2Fed8c715b-81ee-4bef-9f46-1e057f6a1313%2FNationalTelecommunicationsAndInformationAdministration_Logo_16x9.png?w=480&auto=format%2Ccompress&fit=max)
The National Telecommunications and Information Administration (NTIA) is seeking inputs on how widely available access to open-weight models may impact national security.
In a recent panel discussion, Alan Davidson, Assistant Secretary of Commerce for Communications and Information and NTIA Administrator, said, “One thing we have already learned is the importance of focusing on the marginal or differential risks and benefits of open weights. For example, we need to measure the risks of open-weight models relative to the risks that already exist today from widely-available information, or from closed models.”
Last month, NTIA also launched a ‘Request for Comment’ on the risks, and benefits associated with advanced artificial intelligence (AI) models with widely available model weights.
Davidson highlighted as part of the feedback that NTIA is also seeking inputs on ways in which open-weight models could support U.S. national security interests.
Moreover, NTIA seeks inputs on “How might foreign adversaries use open-weight models to exacerbate security risks? How do those marginal risks compare with closed models?”
“As you can see, we are digging into this topic – and we have a tight deadline for doing so. We hope you all will participate in our comment proceeding – due by March 27,” Davidson said.
Secretary of Commerce Gina Raimondo was also part of the panel discussions. She discussed the incremental advantages and drawbacks associated with open-weight AI models, emphasizing the importance of accountability systems and risk mitigation strategies.
Additionally, she touched upon the national security and societal implications of open AI models, as well as current and evolving methods for evaluating AI technologies.
Attendees included representatives from AI labs, civil society organizations, and academics such as the Center for Democracy & Technology, Anthropic, Google DeepMind, Greylock, Meta, MLCommons, OpenAI, RAND, Scale AI, the Stanford Center for Research on Foundation Models, and Upturn.