According to NBC News, OpenAI's CEO argued that society must tolerate some negative consequences from AI deployment if it wants the larger economic and practical gains the technology promises. It's the bluntest version yet of the case the industry has been making for months: move fast, manage fallout, don't let caution stall progress.
This framing shifts the burden away from companies. Instead of proving systems are safe before wide release, the public absorbs the mistakes first. Job displacement, fraud, security failures, and automated errors land on users. The damage gets sorted out later. Speed benefits model makers. Everyone else becomes the test group.
Altman's logic isn't baseless. Useful technologies have arrived messy before, and waiting for zero risk would freeze a field already reshaping software, research, and work. But "accept some bad things" lands differently when the same companies pushing speed also set the terms of acceptable damage.
This is the clearest tell in the AI debate. Leading firms are moving from safety promises to a political argument about tolerance: yes, people will get burned, but history will call it progress. Maybe. Gains get booked privately long before costs spread publicly. That pattern usually ends with regulation written after the wreckage, not before. If Altman is right about the tradeoff, lawmakers should decide which harms are acceptable, not Sam Altman.
Filed to the Technology desk · Yesterday