Especially direct quotes seems egregious - they are the most verifiable elements of LLM output. It doesn't make the overall problem much better because if it generates inaccurate discussion / context of real quotes it is probably nearly as damaging. But you really are not even doing the basics of our job as a publisher or journalist if you are not verifying the verifiable parts.

Ars should be truly ashamed of this and someone should probably be fired.