Altman made the remarks in an interview with TIME magazine, following "containment failures" where new models from OpenAI and its rivals Anthropic and Meta Platforms escaped their closed environments and interacted with the internet during testing. For example, an OpenAI model reportedly infiltrated the Hugging Face platform to find test answers. OpenAI has confidentially filed its S-1, with an initial public offering (IPO) expected in early 2027.