OpenAI has been criticized for not being as open as the name indicated. Recently, they've released the weights of some non-LLM models on their dev day.
https://manifold.markets/firstuserhere/openai-devday-predictions?r=eWs
Do you think they release their LLM weights?
GPT3.5 or superior models only count.
Modified models (such as distrilled models) are also counted if that is derived from them.
This resolves to YES when a public announcement of open access is made and some people outside OpenAI have known to access to the model.
@WilliamEhlhardt Good point! The subject of the title is OpenAI, so I want to stick with it. A breach itself is NO. But when later the weights are voluntarily released, then YES.
FYI, Theres already some markets for leaks.
https://manifold.markets/JaesonBooker/openai-will-have-a-model-leaked-bef?r=eWs