> If #2 was false, surely some other LLM lab would have done it by now.

This is a logical flaw. LLM that is immune to jailbreak _could_ exist, but not yet, or maybe nobody talks about it. Yes there's a market, but all of these AI boom is too recent to make any claims.