On one hand, open source might diminish the arms race dynamics playing out in the development of these models. On the other, this significantly increases the complexity and risks associated with these technologies. What AI alignment looks like in such an open environment?