Accepting Some AI Harms

Sam Altman argues that society should tolerate certain non‑catastrophic harms from AI (scams, misuse, smaller hacks) in order to preserve broad public access and individual agency rather than vesting exclusive custody in a single lab or imposing total bans. OpenAI is publicly endorsing targeted state safety laws and outside evaluators while drawing a line at runaway, catastrophic loss of control. — This frames the central regulatory trade‑off — access and agency versus minimizing everyday harms and centralization — and will shape legislative fights, public opinion, and industry coalitions.