Today, AI models are often trained in multi-stakeholder research collaborations among companies, academics, and individual developers. AI models are increasingly shared publicly and hosted on platforms like GitHub and HuggingFace. These models are then often leveraged into development and deployed. Regulations that assume AI R&D is done by a single actor or single type of actor might be poorly fit for purpose. Many countries have introduced exceptions for certain text and data mining or are considering doing so.
• What does AI R&D look like today, who are the actors involved, and how do they collaborate?
• How can we ensure that policymakers get a clearer picture of AI R&D practices?
• How are text and data-mining exemptions to copyright law relevant in AI R&D?
• What are key challenges and opportunities in setting these rules of the road for AI?