Which is why it’s important to be able to run models locally. Which also might explain the strategy behind buying all of the memory that is or will exist for at least a year out. Maybe we’ll eventually see AI safety be used to prevent people from running local models.

Given the recent mexican telecom hacks were allegedly done with significant help from openai/anthropic's chatbots, it seems at least somewhat prudent to require some sort of identity verification for API access? I'm struggling to see how this isn't the tech community's version of "no background checks for gun purchases" or "no KYC for bank accounts".