HUMANLEDGER

The standard · annexes

Sector annexes

The core is the same everywhere, and the annexes show what it looks like in one sector’s P&L. One is written. The other five are open, and they are meant to look it.

The annex rules · agreed 22 Sep 2026

Three rules govern what an annex may contain and who writes the next one.

  1. Core conditions are what everyone discloses. Six conditions, identical in every sector, no substitutions.
  2. An annex metric is a sector-specific rendering of a core condition, meaning the same thing measured in the units that sector actually keeps. Judgement rate renders Exposure at the operator position; assets per reviewer per year renders liability without authority for an agency.
  3. Any metric that cannot be traced back to a core condition does not belong in an annex. This is the traceability test. It would have caught originality distance and trust delta before publication, since neither traces to a condition: both judge the work.

Two working notes that follow from them:

  • Not every cell has to be populated in every sector. An empty harm class is a disclosure rather than an omission: where AI never decides anything about a person, adverse decisions sit near zero, and that is a legitimate result, not a failed annex.
  • Annexes set thresholds; the core mandates that a threshold exists. A universal minimum review time is indefensible across a mortgage decision and a content approval, so the core requires one to be set and published per decision class, and the annex supplies the value.

Which annex gets written next is decided by evidence, not appetite: the sector accumulating the most signals in the log. See the signals, tagged by sector →

All six annexes

Annex A · Written

Creative, marketing and media

Agencies, brands, studios and publishers: the canary rather than the subject.

Read Annex A → The table →

The structure is borrowed from the GHG Protocol: one universal core that nobody gets to rewrite for their own sector, and sector guidance on top of it. A universal standard with no worked sector is a manifesto; a universal standard with one deeply worked sector is a spec.

Annex A is creative, marketing and media because that is where AI landed first and hardest, and it is the sector the author can write about with authority. That makes it the first worked example, not the centre of the standard. The rest need people who work in them, and the race between them is visible: the signals log tags every entry by sector.