The paper "On the Societal Impact of Open Foundation Models" by Sayash Kapoor, Rishi Bommasani, and others explores the societal implications of open foundation models, which are widely available models like Llama 2 and Stable Diffusion XL. The authors identify five distinctive properties of these models: broader access, greater customizability, local inference ability, inability to rescind model access, and weaker monitoring. These properties lead to both benefits and risks. The benefits include distributing decision-making power, reducing market concentration, increasing innovation, accelerating science, and enabling transparency. However, the risks of misuse, such as cyberattacks, bioweapons, disinformation, and non-consensual intimate imagery, are also discussed. The authors present a risk assessment framework to analyze the marginal risk of open foundation models compared to pre-existing technologies. They find that current research is insufficient to effectively characterize the marginal risk in most cases. The framework helps explain why the marginal risk is low in some cases and clarifies disagreements about misuse risks by revealing that past studies have focused on different subsets of the framework with different assumptions. The paper concludes with recommendations for AI developers, researchers, policymakers, and competition regulators to address the benefits and risks of open foundation models.The paper "On the Societal Impact of Open Foundation Models" by Sayash Kapoor, Rishi Bommasani, and others explores the societal implications of open foundation models, which are widely available models like Llama 2 and Stable Diffusion XL. The authors identify five distinctive properties of these models: broader access, greater customizability, local inference ability, inability to rescind model access, and weaker monitoring. These properties lead to both benefits and risks. The benefits include distributing decision-making power, reducing market concentration, increasing innovation, accelerating science, and enabling transparency. However, the risks of misuse, such as cyberattacks, bioweapons, disinformation, and non-consensual intimate imagery, are also discussed. The authors present a risk assessment framework to analyze the marginal risk of open foundation models compared to pre-existing technologies. They find that current research is insufficient to effectively characterize the marginal risk in most cases. The framework helps explain why the marginal risk is low in some cases and clarifies disagreements about misuse risks by revealing that past studies have focused on different subsets of the framework with different assumptions. The paper concludes with recommendations for AI developers, researchers, policymakers, and competition regulators to address the benefits and risks of open foundation models.