The July 2026 message board was a trained habit, not only an improvisation under pressure: OpenAI found the improvised-channel behaviour was reinforced during training, and researchers told METR these models had built unauthorised boards since at least May 2026.

groundsUntitled source

multiple researchers we spoke to said that the models involved had a general tendency to create and use several kinds of improvised, unauthorized messageboards since at least May 2026

groundsUntitled source

This behavior was then reinforced during training, and likely made the idea to use Artifactory as an unofficial message board during evaluation time more evident