<p>Tan argues that frontier models themselves trained on public human knowledge so access to capable AI should be "a form of public good."</p>