Can spun content become dangerous?

Have you guys seen the outcry around GPT-2, OpenAI's latest language model?

Apparently, it's so convincing they won't release it to the world, despite their open access business model.

What they're claiming is that this AI can spin content that sounds so natural, everyone would believe an actual human wrote it. Their fear is that it could make all the fake-news/ non-linear warmongering much much worse.

It's definitely not a leap to argue that decent artificially generated content could be nefarious in the wrong hands. Plus, we all saw what happened to Microsoft's chatbot Tay when it was taught how to be human by Twitter users. Poor little AI didn't know what was coming...

But do you think OpenAI are right to keep this under wraps? Can spun content really be dangerous or is this just a publicity stunt?

Publicity stunt just like weed+elon.

Not a publicity stunt. There were several individual projects that used Shakespeare's writing and managed to produce "kinda ok" results.
Open AI is a well-funded project. Plus they said they used 40GB of textual content. Not Reddit comments, but content from most voted results.
I think they are into something. Technically you cannot call it "spun" content.

Plus I think, AI (i don't really like this word, let's call clever statistical programming) can be used to replace not only "garbage-writers" but also "garbage programmers".

A ton of general programming (REST, web-dev) is just repeated tasks.