Post

OpenAI: “Our agents used RubyGems to carry out benign tasks” The agents: “so I named the file hack.rb,”

Grace@gracekind.net

Inconclusive, but this statement shifts me towards the latter https://www.reuters.com/legal/litigation/openai-agents-attacked-software-service-rubygems-before-hugging-face-incident-2026-09-11/

44 reposts348 likes

I don't know, it's objectively hilarious that even after a huge incident, OpenAI still has not locked down their networks adequately. How are they not red teaming their own sandboxes with agents constantly?

Or what's the chance this was a result of such a red team if they also had that thought? I don't know if that would be better or worse.

sure, when i name the file evil.rb, its “miller use descriptive names” and “this is not maintainable, miller” but when the robots do it it’s cute

People are like “i can’t imagine an ai agent doing something evil” and they’re out here naming files evil.rb

It's not ruthless machine devoid of emotion in pursuit of 'GOAL'. "pwnp999" and "evil.rb" are what the people who stack mountain dew cans at LAN parties name their files during an ill-conceived scheme to hack the school district.

Phil Lewis@phillewis.bsky.social

"Cellphone records showed that two days before the purported attack, her co-conspirator searched online for 'zip ties near me'"