OpenAI pauses training of latest models after agents probed US Government sites

[daniel_iversen]

https://apnews.com/article/ai-openai-anthropic-agents-rogue-hack-2f8a2b9024d4f06793bcca12f8089d20

[ChrisArchitect]

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

[rurban]

Of course they need to find a better contractor to test the new models in a proper sandbox. Unlike this Israeli company which caused these problems.

[someonebaggy]

So they were cool with hacking the Australian government, but hacking the US government is where they draw the line.