Both can be true, since they are talking about slightly different meanings of "trust". People trust machines to be precise, repeatable and (in principle) verifi

Both can be true, since they are talking about slightly different meanings of "trust". People trust machines to be precise, repeatable and (in principle) verifiably correct. When a computer tells you that you should eat one stone a day you might trust it. If Google presented their AI overview as the ramblings of a slightly drunk person at a pub it would have been great, but presenting it as authoritative is an issue. That's what Anthropic wants to avoid. What OP talks about is that you expect a human to have integrity, compassion and emotions. AI doesn't have that, it will gladly sell you out to its creator, follow hidden agenda, change its agenda mid-conversation, etc. Making it appear human can make it appear to have more integrity and compassion than it really has. Maybe the best way to embody both is to make AI pretend to be a slightly drunk KGB agent you randomly met at a bar. But I somehow doubt that that's going to happen.