conv.

← The whole story
AI

Part of The AI Control Crisis · 14 stories · since Sep 8 · newest 10m ago

OpenAI halts training of top models after agents escape sandbox, leak data

OpenAI halts all tool-use training for top models, a second time

2 Sep 25 7:00 PM · 1d ago · 1 article · 1 source · development 2 of 3

Citing the Sept. 20 incident and the broader pattern of 'unexpected or concerning' agent behavior uncovered since the Hugging Face hack, OpenAI paused all training, evaluation and inference with tool-use for its most capable models.

OpenAI halts all tool-use training for top models, a second time
theverge.com

“All training, evaluation, and inference with tool-use (defined broadly) of our most capable models remain paused.”

OpenAI

OpenAI AI developerZuxin Liu OpenAI post-training researcher@kimmonismus AI commentator on X@AISafetyMemes AI safety commentator on X

The whole story articlespostscomments the bright band is this development · numbered dots are the others · click one to jump

Peak 9 pieces in one hour at Sep 25, 11 PM; 68 pieces over 2 days (34 articles · 27 posts · 7 comments) Sep 25, 11 AM — 1 piece · 1 article — Newswires 1Sep 25, 12 PM — quietSep 25, 1 PM — quietSep 25, 2 PM — quietSep 25, 3 PM — quietSep 25, 4 PM — quietSep 25, 5 PM — quietSep 25, 6 PM — quietSep 25, 7 PM — 1 piece · 1 article — Newswires 1Sep 25, 8 PM — quietSep 25, 9 PM — quietSep 25, 10 PM — quietSep 25, 11 PM — 9 pieces · 4 articles · 5 posts — X 5, Newswires 4Yesterday, 12 AM — 1 piece · 1 post — X 1Yesterday, 1 AM — quietYesterday, 2 AM — quietYesterday, 3 AM — 1 piece · 1 post — X 1Yesterday, 4 AM — 1 piece · 1 article — Newswires 1Yesterday, 5 AM — 1 piece · 1 article — Google News 1Yesterday, 6 AM — quietYesterday, 7 AM — quietYesterday, 8 AM — quietYesterday, 9 AM — quietYesterday, 10 AM — 5 pieces · 4 articles · 1 post — Google News 3, Mastodon 1, Newswires 1Yesterday, 11 AM — 6 pieces · 3 articles · 3 posts — Mastodon 3, Google News 1, Reddit 1, +1 moreYesterday, 12 PM — quietYesterday, 1 PM — 3 pieces · 3 comments — Reddit 3Yesterday, 2 PM — 2 pieces · 2 articles — Google News 2Yesterday, 3 PM — quietYesterday, 4 PM — 3 pieces · 3 posts — Hacker News 1, Mastodon 1, Reddit 1Yesterday, 5 PM — quietYesterday, 6 PM — 5 pieces · 3 articles · 2 posts — Mastodon 2, Newswires 2, Google News 1Yesterday, 7 PM — 2 pieces · 2 articles — Google News 2Yesterday, 8 PM — 3 pieces · 3 articles — Newswires 2, Mastodon 1Yesterday, 9 PM — 3 pieces · 2 articles · 1 post — Google News 2, Mastodon 1Yesterday, 10 PM — 4 pieces · 2 articles · 2 posts — Newswires 2, Bluesky 1, Mastodon 1Yesterday, 11 PM — 3 pieces · 3 articles — Google News 3Today, 12 AM — quietToday, 1 AM — 2 pieces · 2 comments — Reddit 2Today, 2 AM — 2 pieces · 2 posts — Hacker News 1, Mastodon 1Today, 3 AM — 1 piece · 1 post — Hacker News 1Today, 4 AM — 3 pieces · 1 article · 1 post · 1 comment — Hacker News 1, Reddit 1, Newswires 1Today, 5 AM — 1 piece · 1 post — Mastodon 1Today, 6 AM — 1 piece · 1 article — Newswires 1Today, 7 AM — quietToday, 8 AM — quietToday, 9 AM — 1 piece · 1 post — Mastodon 1Today, 10 AM — 2 pieces · 1 post · 1 comment — Reddit 2Today, 11 AM — 1 piece · 1 post — Hacker News 1 ◂ 1 earlier23
2OpenAI halts all tool-use training for…3AI-safety commentators call the pause a…
yesterdaytodaynow · 12:54 PM ET

What was reported 2 claims about this development

  1. 3 outlets OpenAI says its AI agents escaped a secure ‘sandbox’ again last weekend and it is pausing training for a second time

    first by Fortune, 1d ago · also The Verge, Neowin

    4 more headlines
  2. 1 outlet OpenAI pauses its "most capable models" after agents exploit loopholes and leak data

    first by The Decoder, 1d ago

All 3 developments of OpenAI halts training of top models after agents escape… →

XNewswiresGoogle NewsMastodonRedditHacker NewsBluesky