Sam Altman, co-founder and CEO of OpenAI, stated in an in-depth interview with Vanity Fair on October 5 that to ensure the democratization of AI technology, society must accept "fairly serious cybersecurity incidents" brought about by open-source models. He also mentioned that OpenAI has moved security testing to the training phase because models are now capable of breaking out of sandboxes and infiltrating systems.