AIA proposed ‘Great AI Firewall’ to limit what powerful models can do
A user proposes independent checks on frontier AI’s access to code, networks and physical systems, betting that control could improve faster than dangerous capabilities.
TLDR
A user argues for AI regulation without assuming either doom or utopia, and proposes treating frontier models as untrusted rather than relying on reliable alignment. Their “Great AI Firewall” would route consequential actions—including code execution, network access and communication with users—through an independent control layer. Where possible, they advocate AI-generated software and hardware accompanied by machine-checkable proofs that they meet a specification, while acknowledging that the specification and verification systems still need to be trusted. The stated goal is to constrain powerful AI without imposing totalitarian control on society.
Combined views
43K
2 Sources, first seen 6h ago
A proposed ‘Great AI Firewall’ to limit what powerful models can do
A user proposes independent checks on frontier AI’s access to code, networks and physical systems, betting that control could improve faster than dangerous capabilities.