Garry Tan Questions Anthropic Compliance With AI Model Sharing Executive Order

Fable 5 is fire and losing capability when you rely on frontier models as much as I do feels like losing a limb

Someone please fill out their TPS reports next time so we don’t have to experience this

@garrytan UK AISI is a lot bigger (and for once better funded) than its US counterpart…

Actually, the US Gov was deeply in the loop from the start. Anthropic included them in Project Glasswing along with Apple and CrowdStrike. The real twist is that the government saw what Mythos could do, panicked over potential jailbreaks, and Trump immediately banned federal agencies from using Anthropic tech altogether. 🏛️🚨

@garrytan sounds like the feds got ghosted again someone forgot to cc washington on that one glasswing flying solo while dc sleeps

@garrytan Plenty of people in the govt have been using it - nothing about the capability of Mythos is new to them

@garrytan Microsoft has delivered “security risk” code for decades; don’t recall government intervention in those cases. Hmmmmmm

@garrytan I'd rather lose that limb than be subject to strict AI KYC requirements. We are headed into a dystopian direction...

It's absolutely sad that it's forced the smartest talent in the US to cheer on Chinese AI labs, we've lost the plot.

@garrytan Like wtf 😂