Sam Altman's October 2026 position, that society should accept limited AI harm in exchange for broader benefits and access, is a more direct articulation of risk tolerance than most AI executives have been willing to state publicly. Where competitors tend to emphasize safety work without specifying what level of harm they consider acceptable, Altman named it explicitly: scams and hacks, yes; catastrophic risk, no.
That specificity cuts both ways for OpenAI. It gives the company a clearer, more defensible public position than vaguer safety commitments, but it also gives critics a concrete claim to hold the company accountable to the next time a serious AI-enabled harm actually occurs and the public asks whether it crossed the line Altman himself drew.