This question is oddly worded, such that it seems meant to elicit only answers about dishonesty, rather than more nuanced takes on the balance of honesty and dishonesty in recruiting.

When I went through a series of interviews with many organizations in 2018, I mostly remember it feeling really honest:

• I had applied for a position as Stuart Russell's personal assistant. When I spoke to him about the role, he frankly told me that he wasn't sure the position would work out at all, and that past personal assistants had done very little to boost his productivity.
• I applied for an ops position at MIRI, and as far as I recall, they made very little effort to "sell me" on the job, even in the final round — we talked about what it would entail, I did an in-office work test, and we were done.
• I applied to AI Impacts and talked to Katja Grace for a while. She was frank about her uncertainty around the impact of the org's research, and was clear with me that my work would be nebulous and that there wasn't much in the way of shovel-ready opportunities that would obviously do a lot of good for someone.
• In one case, when an org called me to tell me about an offer, I told them I was already considering an offer from a different org. Their response was something like "nice, that also sounds like a great opportunity" (rather than the "nah, you'll do better here, come work for us" that I'd expect in the private sector).

Not all the places I applied to had conspicuous moments of unusual honesty like this, but I don't remember feeling like anyone had given me a hard sell.

Now that I'm in the midst of recruiting a new content specialist, I'm trying to avoid impact bias by:

• Explaining to candidates how often we end up feeling really uncertain about our counterfactual impact
• Telling candidates exactly how many people are left in the process at each stage, so that they get a sense for how many people are around the same level of fit as they are
• Portraying the role as very flexible and malleable, such that many different people (or sets of multiple people) could cover it well (rather than the role requiring very particular skills that just so happen to be the same as the candidate's)

My answer: 2