AI safety needs more than slogans — it needs a reliable kill switch.

At North River Labs, we’ve been working on exactly that: a hardware-rooted kill switch that ties every AI system to a confirmed physical device identity. Instead of begging a misaligned agent to shut itself down, our Veris platform cuts its identity out of the network — revoking its cryptographic credentials so it can’t authenticate, move laterally, or reach critical systems, even if the model keeps running on a dark island.

With more than a billion autonomous AI agents projected by 2028, governance has to start at the device layer — not in policy docs, but in the physics and cryptography underneath the stack.