OpenAI has published the text-generating AI previously labeled ‘too dangerous’ to share

Research lab OpenAI has now released in full a text-generating AI system that many had warned could be used maliciously. While the lab had announced the system GPT-2, in February 2019, it did not release the full version citing fears about it being used to generate fake news, spam and other kinds of disinformation.

Now that it has been released in full, OpenAI has said in a blog post that it has so far seen ‘no strong evidence of misuse’. GPT-2 has impressed many as being part of a new variety of text-generating systems which are able to generate intelligible text from minimal prompts. The system is said to have been trained on 8 million text documents scraped from the web.

You can try a web version at and enter your own prompts to see the results.

Sign Up for nextbigwhat newsletter

The smartest newsletter, partly written by AI.

Download, the short news app for busy professionals