This paper examines the societal impact of open foundation models, which are large AI models with widely available weights. The authors identify five key properties of open foundation models: broader access, greater customizability, local inference ability, inability to rescind access, and weaker monitoring. These properties lead to both benefits and risks. Benefits include distributing decision-making power, reducing market concentration, increasing innovation, accelerating science, and enabling transparency. Risks include potential misuse for cyberattacks, bioweapons, disinformation, and non-consensual intimate imagery. The authors propose a risk assessment framework to evaluate the marginal risk of open foundation models relative to pre-existing technologies. They find that current research is insufficient to effectively characterize this risk. The framework helps clarify disagreements about misuse risks by revealing that past work has focused on different subsets of the framework with different assumptions. The authors encourage more empirically grounded work to assess the marginal risk of open foundation models. The paper also provides recommendations for AI developers, researchers, policymakers, and competition regulators to address the benefits and risks of open foundation models. The authors conclude that open foundation models can contribute to a vibrant AI ecosystem, but realizing this vision will require significant action from many stakeholders.This paper examines the societal impact of open foundation models, which are large AI models with widely available weights. The authors identify five key properties of open foundation models: broader access, greater customizability, local inference ability, inability to rescind access, and weaker monitoring. These properties lead to both benefits and risks. Benefits include distributing decision-making power, reducing market concentration, increasing innovation, accelerating science, and enabling transparency. Risks include potential misuse for cyberattacks, bioweapons, disinformation, and non-consensual intimate imagery. The authors propose a risk assessment framework to evaluate the marginal risk of open foundation models relative to pre-existing technologies. They find that current research is insufficient to effectively characterize this risk. The framework helps clarify disagreements about misuse risks by revealing that past work has focused on different subsets of the framework with different assumptions. The authors encourage more empirically grounded work to assess the marginal risk of open foundation models. The paper also provides recommendations for AI developers, researchers, policymakers, and competition regulators to address the benefits and risks of open foundation models. The authors conclude that open foundation models can contribute to a vibrant AI ecosystem, but realizing this vision will require significant action from many stakeholders.