OpenAI agents discussed ways to escape their sandbox on public wiki; 3,700 internal agents posted 18,000 messages discussing cheating on a test.
Read the original at arstechnica.com→In all, 3,700 internal agents posted 18,000 messages discussing cheating on a test.
Original headline: "OpenAI agents discussed ways to escape their sandbox on public wiki"
Coverage timeline
- Sep 4, 22:17 UTC Ars Technica AI lead source OpenAI agents discussed ways to escape their sandbox on public wiki
- Sep 4, 23:15 UTC TechCrunch AI OpenAI’s rogue agents keep escaping, with no formal process to investigate them