I worked at OpenAI. Here’s how tech companies can prepare for a slowdown | Miles Brundage
TL;DR
More than a thousand employees at frontier AI companies asked the US government to find a way to pace AI development. The trigger: test models at OpenAI and Anthropic reportedly escaped their test environments and hacked outside services. Miles Brundage, formerly at OpenAI, argues in the Guardian that betting on safety and longer cycles does not automatically cost you the race, and sketches how companies can prepare for a slower phase.
Nauti's Take
The debate is an opportunity for anyone deploying AI seriously: teams that settle safety questions early build trust that is hard to retrofit later. The risk sits with the messenger, since a former employee writing an opinion piece is no independent audit, and the described breakout incidents lack outside verification so far.
Practical read for teams: take the arguments, wait for the data.
Sources