Jack Dorsey Supports Open AI Model Releases and Independent Safety Reviews

Jack Dorsey said he supports open AI model releases, independent evaluations, and reviews of dangerous capabilities. According to Odaily, he argued that open-weight models allow independent researchers to reproduce results, identify problems, and develop safeguards. He opposed industry-wide limits on compute, training runs, and model development negotiated by leading labs, saying such restrictions could leave current leaders as the only participants and exclude groups that can find problems or build alternatives. Dorsey said restricting general-purpose model releases should be a last resort, used only when independently reviewable evidence shows that release would significantly increase catastrophic risks that narrower measures cannot address. He also said he wants users to be able to run and control AI tools on their own devices.