U.S. Commerce Department Supports Open-Weight Generative AI Models, Calls for Monitoring and Risk Evaluation
The U.S. Commerce Department’s National Telecommunications and Information Administration (NTIA) has released a report endorsing the use of “open-weight” generative AI models like Meta’s Llama 3.1. These models, according to the report, promote wider availability of generative AI for small companies, researchers, nonprofits, and individual developers. The report suggests that the government should not impose restrictions on access to open models without first investigating whether such restrictions may harm the market.
This viewpoint is in line with recent statements made by FTC Commission chair Lina Khan, who believes that open models can foster healthy competition by allowing more small players to bring their ideas to market.
Alan Davidson, assistant secretary of Commerce for Communications and Information and NTIA administrator, emphasizes the importance of open AI systems in promoting competition, innovation, and managing risks. He acknowledges that the government has a crucial role to play in supporting AI development while actively monitoring the risks associated with the wide availability of model weights for large AI models.
As regulators both domestically and internationally consider new rules and requirements for companies releasing open-weight models, the NTIA’s report provides valuable insights. California is on the verge of passing SB 1047, a bill that would mandate stronger cybersecurity measures and the capability to “shut down” copies of models for companies training AI models using more than 1026 FLOP of compute power. The EU has also finalized compliance deadlines under its AI Act, introducing rules concerning copyright, transparency, and AI applications.
Meta has expressed concerns that the EU’s AI policies will restrict its ability to release certain open models in the future. Additionally, several startups and big tech companies have criticized California’s law, deeming it overly burdensome.
While the NTIA’s report supports the availability of open models, it also acknowledges the need for model governance. The report suggests that the government should establish an ongoing program to collect evidence on the risks and benefits of open models, evaluate that evidence, and take appropriate action, including imposing restrictions on model availability if necessary. The report proposes that the government invest in researching the safety of different AI models, support risk mitigation research, and develop risk-specific indicators to signal the need for policy changes.
These proposed steps align with President Joe Biden’s executive order on AI, which calls for the development of new standards by government agencies and companies in the creation, deployment, and use of AI. U.S. Secretary of Commerce Gina Raimondo emphasizes the Biden administration’s commitment to maximizing the potential of AI while minimizing its risks. She states that the report provides a roadmap for responsible AI innovation and American leadership by embracing openness and offering recommendations on how the U.S. government can prepare for and address potential challenges in the future.