← back to the lobby

Safety and technology, inseparable

VERIFIABLEETHICSVoiceOversightLimitsHaltKEYSTONEPrivacyDignityLifecycleL5ApplicationsL4GovernanceL3GrammarL2The EnvelopeL1Transport
Block 01 · Halt◆ KEYSTONE
The kill switch

There is an off-switch that no score, no argument, and no attacker can override. Remove this one block and nothing else can hold the load.

Why it cannot be removed
01
Live-quorum halt
HUMANITY_ACCORD: 2-of-3 named humans, live-quorum so it works even if the roster is attacked or comms are jammed.
02
Non-revocable
No party inside the federation can disable or route around the stop.
03
Above every score
The halt sits over the whole system. No objective can outvote it.
The keystone. Every other guarantee presses against it, and it is the one block that cannot be removed.
Explore The kill switch
The ground it springs from
Independently verifiable — every block is checkable by parties the AI does not control, and a small primitive surface shrinks the attack surface.
Legally aligned — GDPR, CCPA, and the EU AI Act map onto the protocol through the DP-Map, with SI-score liability and sector overlays.
Why the span holds: the coherence ratchet

An arch grows stronger the more weight presses on it. The load locks the stones tighter. CIRIS holds the same way. Every voice is priced honestly: copies of one voice count as one voice, and no amount of echo can buy more trust. The books stay true, signed and tamper-evident, capped by the halt. When something fails, it fails in the record, under its own name.

CIRIS · safe by structure, open by principle, kind by design