The pace of AI model releases is accelerating, forcing safety evaluators to shift from pre-deployment to post-deployment research.

A government-run AI regulator would rapidly create a testing queue starting with a month-long delay and growing to many months, causing the US to lose the AI race.

“You're going to very rapidly end up with a queue where all the models would be waiting to get tested and it would start with a month-long delay. It would end up being many months and we would just lose the AI race.”

UK AISI is shifting from pre-deployment evaluations to longer post-deployment research collaborations due to the increasing pace of model releases.

“We are shifting a lot of that work not least because the pace of model releases is increasing to longer research collaborations that might go back either are post deployment or they go back further before deployment is finalized.”

The US government got involved in the release of Anthropic's Mythos model and extended its delay almost into July.

“Also the government got involved, so it ended up being extended almost into July.”