Exploring the Implications of Open AI Model Weights

OpenAI’s Advocacy for Innovation and Risk Mitigation

OpenAI submitted a comment to the National Telecommunications and Information Administration (NTIA) regarding the release of open model weights. In their comment, OpenAI highlighted the importance of fostering innovation and competition in the AI sector through the release of model weights. They have previously released some of their models’ weights and developed infrastructure to support other AI developers. OpenAI advocates for a balanced approach that includes both open weights releases and API/product-based releases to study and mitigate risks associated with AI technologies. They emphasize the need for iterative deployment and the development of a Preparedness Framework to assess and mitigate catastrophic risks associated with AI. OpenAI suggests that assessments for catastrophic risks may not be necessary for models that are less resource-intensive. They call for the development of a better science of AI risks and support government involvement in risk evaluation practices. OpenAI stresses the importance of flexible government policies that can adapt to future changes in the AI landscape (OpenAI’s comment).

NTIA’s Request for Information on Dual-Use Foundation Models

The NTIA issued a Request for Information (RFI) to gather public input on the potential benefits, risks, and implications of dual-use foundation models for which the model weights are widely available. The RFI aims to understand the definition of “widely available,” the potential risks to security and equity, and the benefits to competition and innovation. The NTIA is also interested in governance, international collaboration, and decision-making processes for determining access to model weights. They are considering the changing technology landscape and potential long-term risks associated with widely available model weights. The NTIA is open to exploring other issues and advancements related to dual-use foundation models (NTIA RFI).

This research indicates that there is a significant interest from both OpenAI and the NTIA in exploring the implications of making AI model weights widely available. OpenAI’s comment to the NTIA underscores the potential for innovation and competition, while also acknowledging the need for careful consideration of risks and the development of frameworks to mitigate those risks. The NTIA’s RFI further emphasizes the government’s interest in understanding the broader implications of open model weights, including security, equity, and governance issues.