A model that didn‘t break out of its sandbox and hacked other websites cannot be good.
To be fair, Google has a real security team.
Or maybe Google actually knows how to set up a VM.
Argon has been hacking Google internally for years!
Related funny https://www.youtube.com/watch?v=DlTNN0gvkLM
They just don't know yet, wait for a few months
People complain about the slop-release. Who knows, maybe it is the first self-release by an AI model in history.
https://www.theguardian.com/technology/2026/sep/18/google-ge...
?