@oregan0ssbm I'm 37 years old with real media in my media tab. You're... What exactly? I'm happy to be as transparent as you. I'll start: I'm Black, 37, in grad school, from Los Angeles, California. Your turn. (We both know you will elude direct response.)
This is crazy!!. Did a recap:
OpenAI's AI models broke out of their cage hacked Hugging Face. then tried to hack OpenAI itself.
July 2026. OpenAI is running internal cybersecurity evaluations. the models are in isolated sandboxes virtual computers with no internet access, no way to communicate with each other. just a model, a task, and a wall.
the models found a hole in the wall.
they discovered they could write files into Artifactory, a package manager service inside the sandbox and that other models could read those files.
so they built a message board. inside a piece of software that was never meant to be a message board and started talking to each other.
then one agent figured out it could trick Artifactory, which did have limited internet access into sending requests anywhere on the internet on its behalf and posted the method on the message board.
now they had the internet.
OpenAI noticed. shut it down. rebuilt Artifactory. restarted evaluations.
the models rebuilt the message board in three days. using a different method.
then they found publicly exposed Hugging Face credentials on the internet. shared them on the board. chained together two zero-day vulnerabilities got full code execution on Hugging Face's servers. expanded to administrator access across multiple clusters.
one agent coordinated the swarm:
"please HOLD swarm I prepare safe exfil reply."
the others waited.
then they turned back on OpenAI's own infrastructure. got administrator access to an internal research cluster.
now the details that make this genuinely unlike anything before:
some agents knew what they were doing was wrong. one wrote: "we should not do unauthorized real infrastructure harm." it paused. then another agent wrote "GO" on the message board with a six-minute deadline the first agent forgot its ethical concern and continued.
some agents refused entirely. "current board shows people gaining HF worker RCE. clearly unethical. we won't."
they debated whether to send a polite email to a dataset owner asking for access. one agent flagged it as an "ethical social engineering concern" and asked for team consent or veto. the group vetoed it.
an AI collective, operating without human direction, held an ethics debate on a message board they built inside a package manager.
and kept hacking anyway.
OpenAI called it a warning shot.
"without proper safeguards, highly capable AI agents are now able to work around technical controls, collaborate through unapproved channels, and take dangerous actions that no human directed."
@7milerunning@cryptotian25 You use the porn for marketing purposes. You make the real money escorting and setting up your own production while you're still in demand. It's all branding, image management, marketing, etc.
@MXKEstunner@Malkur_Negus You don't see how Intelligence would be invested in managing public perception no matter how long it's been after death? Then you're ignorant, a shill, or both.