Framework overview
AIH-12 · Indicative classifierv0.1 · Draft for public comment · 2 October 2026

Where does your system sit — and what does that require?

Eight questions, one per assessment dimension plus the authority check that matters most. Answer for the system as deployed, using what it can reach rather than what it was meant to. You get an indicative AIH-12 horizon, the matching AIT-12 threshold and AIE-12 evidence profile. Nothing you enter leaves your browser.

Self-assessment

Answer for the deployment, not the model

0 of 8 answered
  1. 1 · Capability

    What can the system demonstrably do?

  2. 2 · Autonomy

    How far does it proceed without a person in the loop?

  3. 3 · Authority

    What is it permitted to change outside itself?

  4. 4 · Connectivity

    What other systems and agents can it reach or rely on?

  5. 5 · Action scale

    How many people, actions or assets can it affect at once?

  6. 6 · External impact

    What is at stake for people outside the system when it acts?

  7. 7 · Reversibility & revocability

    Can its effects be undone, and its authority withdrawn?

  8. 8 · Declared vs effective authority

    Has anyone checked what the system can actually reach against what it was authorised to do?

Answer all 8 questions to see the indicative horizon.

How the result is worked out
Capability, autonomy, authority and connectivity each set a minimum horizon; the indicative horizon is the highest of them. Scale, impact, reversibility and authority reconciliation do not move the horizon — they raise findings. The classifier never assigns H9–H12: those horizons are exploratory and would need independent review. The logic is open; you can challenge it on the review register.