US Senator Cynthia Lummis has put forward the Responsible Innovation and Safe Expertise Act, a proposed law that seeks to provide clarity on liability for artificial intelligence used in professional settings. The bill promotes transparency among AI developers without necessitating open-source models. According to Senator Lummis, the RISE Act ensures that professionals, such as doctors and lawyers, remain accountable for the advice they provide, even when it is informed by AI systems. To avoid civil liability, AI developers would need to publicly release detailed technical documents, known as model cards, which outline an AI system's training data, intended use, performance metrics, limitations, and potential failure modes.
This information is designed to help professionals assess the suitability of AI tools for their work. The proposed legislation does not provide blanket immunity for AI developers but instead sets clear boundaries, excluding protection in cases of recklessness or misconduct. Developers would also be required to update AI documentation and specifications within 30 days of deploying new versions or discovering significant issues, ensuring ongoing transparency.
While the RISE Act stops short of requiring AI models to be fully open-source, it does allow developers to withhold proprietary information, provided that the omitted material is not related to safety and is accompanied by a written justification.