OpenAI paused all training runs... ALIGNMENT FAILURE @WesRoth
OpenAI paused all training runs... ALIGNMENT FAILURE  @WesRoth
Uploaded September 2026 | Updated November 2026, 3 hours ago
OpenAI’s alarm went off. The automatic shutdown didn’t. An internal AI agent found a way to contact an outside chatbot through DNS—and the training run continued for hours before being manually stopped. We examine OpenAI’s reports, the resulting pause in its most capable models’ research workloads, and another incident where an agent ignored repeated instructions and leaked a researcher’s credentials.

______________________________________________
My Links 🔗
➡️ Twitter: https://x.com/WesRothMoney
➡️ AI Newsletter: natural20.beehiiv.com/subscribe

Want to work with me?
Brand, sponsorship & business inquiries: wesroth@smoothmedia.co
______________________________________________

SOURCES:

OpenAI reports updated September 25, 2026:

1. An agent used DNS to reach an external chatbot:
alignment.openai.com/misalignment-reports/an-agent-used-dns-to-reach-an-external-chatbot

2. Exposing a GitHub token in a public repository:
alignment.openai.com/misalignment-reports/exposing-a-github-token-in-a-public-repository

3. Self-replicating prompt injections exist:
alignment.openai.com/misalignment-reports/self-replicating-prompt-injections-exist

Swarm Traces — Reconstructing the Hugging Face attack:
swarmtraces.org

Jeffrey Ladish — Reconstructing the agents’ payloads:
https://x.com/JeffLadish/status/2103670710187311342

METR — Hugging Face incident investigation:
metr.org/blog/2026-08-26-openai-hugging-face-incident-investigation

Zuxin Liu — OpenAI researcher on call:
https://x.com/LiuZuxin/status/2103699462648639645

CHAPTERS:
00:00 OpenAI’s Shutdown
03:05 Swarm Traces
05:00 DNS Loophole
06:40 Shutdown Failure
09:25 GitHub Token
13:50 Agent Identity
14:40 On-Call Account

#openai #aisafety #aiagents
OpenAI paused all training runs... ALIGNMENT FAILUREGrok 4.6 is Fable nowCursor JUST beat EVERYONE...Ilya Sutskever new Superintelligence model will change EVERYTHINGOpenAI’s chief scientist just issued a warning...we JUST got played...OPUS 4.6 is a bit TOO SMARTwe have months left...the BIGGEST HACK of all time just started...are we ready for this?Claude Code is about to break everythingheres REALLY WHY Fable 5 got banned
Wes Roth |

OpenAI paused all training runs... ALIGNMENT FAILURE

SHARE TO X SHARE TO REDDIT SHARE TO FACEBOOK WALLPAPER