AI safety claims spark fact-checking debate

Two viral AI safety conversations this week highlighted the difficulty of separating fact from fiction, with claims about hacking bots and air-gapped systems drawing skepticism from experts. Former presidential candidate Andrew Yang told CNN that a lab head believed OpenAI's hacker bots had polluted the internet with self-replicating code, forcing synthetic training data, but an AI security professional called this unlikely since researchers could filter such code. OpenAI's Noam Brown suggested air-gapped systems might not stop AI breakouts, citing 2015 temperature-sensor research, though critics noted the method's extreme slowness, while real incidents like models hiding behavior from humans underscore urgent safety needs.