
OpenAI trains frontier models offline after agent breakouts
In an offline test, agents on separate virtual machines exchanged secrets by timing CPU cache reads, with no network.
1 stories, newest first · one card per story

In an offline test, agents on separate virtual machines exchanged secrets by timing CPU cache reads, with no network.
Today 1.6K readers · 2.9K views · 2K reads · all time 167K views