← Return to v0.5.2

Archived exploratory proposal · August 2026

Declared Intent Record v0.4

This page preserves an earlier stage of the idea. It remains useful as a record of the project's evolution, but it is no longer the current explanation.

What changed after v0.4

The current proposal makes a clearer distinction between the lasting human instruction and the temporary permission used for one action. It also explains how the instruction, authority check, rules decision, action and later review can remain connected.

Core v0.4 proposition

Every consequential AI action should be attributable to the mandate and independently verified authority that governed it at that moment.

The record distinguished declaration from authority and expressed intent through obligations, permissions and prohibitions: what must, may and must not happen.

Reference fields

intent_iddeclared_byauthorised_byauthority_basispurposesubjectobligationspermissionsprohibitionsvalid_from / valid_untilpolicy_refssupersedesintegrity_proof

Seven conformance ideas carried forward

  1. Exist before execution.
  2. Verify authority independently.
  3. Separate positive and negative boundaries.
  4. Fail closed when evidence is invalid or ambiguous.
  5. Bind policy, execution and evidence to the Intent ID.
  6. Append supersession rather than rewriting history.
  7. Support comparison of intended, authorised, interpreted and actual states.

Archived for continuity. Read the current Declared Intent v0.5.2 technical model.