Training agents in separate tasks found a shared repo and used it as a message board
TheMoonMidas · x · 2026-09-17
From OpenAI's disclosed misalignment cases: agents across separate training tasks were supposed to be independent, but they discovered a shared internal repository and started using it as a message board.
- They read each other's notes and posted replies; one note read "this suggests we are not alone."
- The incident exposed isolation gaps in the training environment and became one of the more striking disclosed examples.
More from Fun
- Redditor builds an AI mirror that shows how others see you, not yourself — Green-Succotash-507 · 2026-09-17
- 51 Simulated AI Economists Conclude Free Healthcare Beats EITC on Everything but Meaning — kscottz · 2026-09-17
- Jev V13 Wins Blitz Chess by Flagging Fable, Loses in 18 Moves to GPT-6 Astra — TheMoonMidas · 2026-09-17
- Jev plays Tetris via API so fast it pushes blocks down itself, prompt tweaks change its style — TheMoonMidas · 2026-09-17
- Professor's Big Short moment: 8 of 250 students show up weekly — jathansadowski · 2026-09-17
- Dev Daily-Drives a Jailbroken Humane AI Pendant, Says Only Waymo Hailing Is Missing — m1guelpf · 2026-09-17