I believe there's a typo in the subtitle of the article:

s/handing of user data/handling of user data/

Just a few months after ChatGPT was released I noticed it just stops the convo in case of asking about suicidal thoughts (I'm okay, I was just investigating)

About half a year ago ChatGPT was generating a new response each time and not showing any phone numbers sometimes at all (not generating them)

So only the last few months they added the standard buttons with real human phone numbers and chats that are likely triggered by simple regex (the AI chat works, too)

So even those visible simple safety things (always adding the standard phone numbers to talk with real humans) that are obvious take them years

And, yep, they were later than Google and Anthropic with this