I believe, having worked a little bit with GPT-2, that OpenAI is intentionally sabotaging access to their AI algorithm and data. They have started this sabotage with GPT-2 and with GPT-3 they simply didn't open source it.

For GPT-2, their repository (https://github.com/openai/gpt-2) is archived. Which is as good as saying that the project is abandoned and will receive no updates. The project already doesn't compile/run properly. This issue (https://github.com/openai/gpt-2/issues/178) could be solved relatively easily by either a one-line code fix, or merging a pull request (https://github.com/openai/gpt-2/pull/244/files). This is not happening and I have a hard time to believe it is on good intentions.

Oh, and by the way, I believe saying "This AI stuff is dangerous to world", the same as politicians saying "We need to check your web history for pedophilia stuff". It's funny how some people don't see the irony in how they oppose one thing and support another that is practically the same.