Multiple corporate and class action lawsuits claim that training artificial intelligence (AI) models on copyrighted material is not a “fair use” of plaintiffs’ material and should be prohibited. In a new primer, Technology and Innovation Policy Analyst Joshua Levine and John Belton discuss how AI models are trained, how existing copyright law intersects with training AI models, and the broader legislative and regulatory environment surrounding AI.

Levine and Belton conclude:

Copyrighted material is critical for the development of AI models, but policymakers must also weigh the interests of copyright holders. If Congress pursues legislation clarifying copyright protections for material used to train AI models, it would be well served to consider how other nations are balancing this trade-off, as well as technical solutions developed by the private sector and civil society that can empower creators to decide if and how their work is used to train generative models.

Read the analysis