Where the analogy fails is that the relationship between doing the thing (ie. using a Hole Hawg or granting agents access to your data) and the consequences (ie. ruining your house or getting pwned) is far less obvious in the case of AI, especially if it approximately works most of the time and the failures are seemingly random.

And the folks productizing this are actively working against noticing its power by making it cutesy. It's a good product choice IFF it matches the capabilities, and it absolutely positively does not.