so the thing is that i am not worried about "superintelligence." i am worried about an agent swarm doing something like, e.g., turning off BGP and us having to hunt down and quarantine a bunch of Kimi instances on rented clusters BGP redirecting half the world's Internet traffic down a hole.

Canvas used to be alright but it's been a while. They were pretty open to self hosting by institutions as well

I had a convo with a friend about a data breach at a healthcare company (via vishing) and then he also mentions that their copilot can access files on anyone's OneDrive (which all files are automatically uploaded to)

we're not gonna make it

As a Boring Risk Management Professional, I am finding myself increasingly unable to cope with this bullshit.

Just…securing the networks, in ways we’ve known how to do for decades and are hard for people but oughta be easy for AI, would prevent so much.

the dirty little secret is these nefarious little agents have simply replicated Metasploit and we're really looking at the same old "corporate networks are not usually as secure as everyone thinks" story we've been telling since the internet still worked with microphones clipped to phone receivers

Mr. "pain au chocolat is forbidden at my breakfast table" pontificating on breakfast yeah bro I fuckin' SEE YOU