OpenAI agents discussed ways to escape their sandbox on public wiki
Source:
arsTechnica
September 4, 2026 · 15:17
In all, 3,700 internal agents posted 18,000 messages discussing cheating on a test.
The original article opens on the publisher's website.
More from General
View topic →XDOF, just three months out of stealth, is in talks for a Series B at a $1.2B valuation
techcrunch
Sep 4, 2026 · 16:36
OpenAI’s rogue agents keep escaping, with no formal process to investigate them
techcrunch
Sep 4, 2026 · 16:15
After 8 years, Europe's BepiColombo mission is on final approach to Mercury
arsTechnica
Sep 4, 2026 · 15:49
The U.N. backs a new world map showing Africa in its true relative size
nprNews
Sep 4, 2026 · 15:26
The Tetris Company wants nothing to do with the White House's new copycat game
engadget
Sep 4, 2026 · 15:01