OpenAI Refuses to Release GPT-2, Fearing Malicious Use

Due to concerns about malicious applications, GPT2 will not be released (2019)

OpenAI trained GPT-2, a 1.5-billion-parameter language model, on 40GB of internet text. It achieves state-of-the-art results on many benchmarks and can perform reading comprehension, translation, and summarization without task-specific training. But because the model could be used to generate misleading news, impersonate people, or automate spam, OpenAI is not releasing the full model or dataset. Instead, they are publishing a smaller version and a technical paper, and will discuss the decision in six months.