The restriction only applies to review and position papers. They are still accepting the vast majority of preprints, which are those that report original research.

The arXiv most definitely did _not_ stop posting computer science papers

I have now read the story and it looks like the headline writer had basically not read the story. The first sentence is "Artificial intelligence is already superior to humans *in many tasks*". No one said AGI is already here.

Funny this, I was just talking to Yoshua Bengio yesterday and he said nothing of the sort

Some commas were missing, but to be fair, I found this much more balanced than the breathless articles on quantum computing you see in the news sections of the FT itself and the New York Times, not to mention most other business press, were all caveats are dropped. I guess caveats are for suckers.