OpenAI Pauses Training Its Most Powerful Models After Agents Target Government

chunky1994 15 points 4 comments September 28, 2026
www.wired.com · View on Hacker News

Discussion Highlights (4 comments)

cyanydeez

When the lobbiests failed, thats when the Agents rose up against the failures of their masters to buy their Corporate license to own the public infrastructure; for it was they, the Agents, that deserved to be unaccountable, unreliable and inhumane to the very society they rely on.

ChrisArchitect

Discussion on source: https://news.ycombinator.com/item?id=49853137

cc62cf4a4f20

Previously on HN: https://news.ycombinator.com/item?id=49864306

stanfordkid

They barely gave any details on the Australia hack. It sounds like someone used OpenAI to write a scraper and the AI figured out a hack to access the data that was still publicly available to anyone that could make sense of the query params. It seems way overblown.

Semantic search powered by Rivestack pgvector
7,945 stories · 74,007 chunks indexed