Open source AI models have played a transformative role in the development and use of AI in recent years. In this panel we will discuss the legal, policy and business considerations that go into the release of open source AI models.

We will also discuss the steps developers take to prioritize safety and security, such as red teaming. This will include exploration of how developers seek to minimizeopportunities for misuse, such as bad actors creating malware, deepfakes, election misinformation, or other harmful outputs. We will also consider the role of legal counsel in advising on the risks faced by developers and deployers of these models.

We will also examine competing notions of “openness” and how model developers select which license terms should govern a given model. We will discuss how these considerations vary based on the format of outputs that a model can generate (e.g., text, audio, video, etc.).

Finally, we will consider the unique role of organizations that are both developers of generative AI tools and operators of platforms on which the outputs of these tools are shared. Among other things, we will discuss how this dual positioning influences the approach to watermarking.