The Milgram experiment is problematic, yes. It's the most narrative-friendly, but there's an enormous body of work on this question, and it seems largely to point in that direction. (Google Scholar claims 819,000 results for "confirmation bias".)

But confirmation bias is very different than believing everything you hear automatically (by default).

You're right.

I get only 5 hits for "default to truth" -- which is so few that I think there must be a better keyword to search for.

Indeed, confirmation bias might be summarized as "hearing what you believe" - very nearly the opposite.