went through the behavioral round at Mistral AI as part of a full loop for an engineering manager role. wanted to document this because the behavioral at AI companies is often treated as an afterthought in prep, and I think that's a mistake here.
it was 45 minutes, one interviewer (an EM or director-level person). felt like a real conversation more than a structured rubric. they had notes, they referenced things I'd said earlier in the loop. not a fresh cold read.
questions I was asked (paraphrased): tell me about a time you pushed back on a direction from leadership and what happened describe a project where you didn't have all the information you needed but had to keep moving. how did you handle uncertainty? what does good technical judgment look like in an environment moving this fast? give me a concrete example. tell me about a decision you made that you'd make differently now how do you think about the safety and societal impact of the things you build?
that last one surprised me. it wasn't a gotcha, they genuinely wanted to hear how I think about it, not a canned answer. I talked about a past situation where I'd flagged concerns about a feature being shipped without sufficient testing. they seemed to actually engage with the substance.
the overall values I sensed them probing: intellectual honesty, comfort with ambiguity, genuine curiosity, and some sense of responsibility about what the company is building.
if I'm being honest: the behavioral round felt more substantive than at some bigger companies where it's pure STAR recitation. they seemed less interested in the format and more in whether you were actually reflective.
prep suggestion: have two or three real moments of disagreement or failure ready. not polished ones. messy ones that you've actually thought about.