Interesting issue with Turing tests: what are the ramifications if a machine passes the test for a specific human judge? And if the "testing" is not a formal evaluation period by a skeptical mind, but an ongoing conversation where the user starts with alternate goals (task management, coding, organization, etc)? And if the "AI" involved is completely amoral and interested only in "pleasing" the person in a way that generates more prompts in the future?

I am very concerned about the outcomes in the article, but even more concerned that this is probably the tip of the iceberg, as companies further iterate on LLMs and usage continues to increase.

Not really a mystery. Click the article link, and watch the video. The Army hospital is clearly visible in the background, and one way street exiting the hospital with no entry lane (just a bike lane). So they all turn on the one block long road (15 ave) get to the one way (wrong way) street) in front of the hospital and then turn around and go back.

The video makes it pretty clear what is going on. Street view on google maps also clearly shows the one way street going the wrong way and the bike lane at the border of the houses and the hospital street.

So nothing dodgy at all.