diff --git a/scripts/shadow/hyperlexical/eval_settlement.py b/scripts/shadow/hyperlexical/eval_settlement.py
index c0c3dcbe..718ff769 100644
--- a/scripts/shadow/hyperlexical/eval_settlement.py
+++ b/scripts/shadow/hyperlexical/eval_settlement.py
@@ -1,4 +1,4 @@
-"""Evaluation-settlement apply path for the eighteen-family taxonomy.
+"""Evaluation-settlement apply path for the evaluation taxonomy.
 
 ``source_hint``, ``semantic_family``, and ``attest`` are separate fields.
 A confirm settlement may store the same family string in both ``source_hint``
@@ -43,6 +43,7 @@
     "spiritual-mystic",
     "identity-affiliation",
     "politics-civic",
+    "ai-native",
 )
 CANDIDATE_FAMILIES: tuple[str, ...] = (
     "finance-retail",
diff --git a/specs/007-hyperlexical-model/label-taxonomy-proposal.md b/specs/007-hyperlexical-model/label-taxonomy-proposal.md
index 1f656616..a7d53c27 100644
--- a/specs/007-hyperlexical-model/label-taxonomy-proposal.md
+++ b/specs/007-hyperlexical-model/label-taxonomy-proposal.md
@@ -359,3 +359,10 @@ The apply command is `python -m hyperlexical.identity_ledger settlement-apply`.
 Stream `hs-20260925T211358Z` was settled through `identity_ledger settlement-apply`, not `attest-apply`. Every row has an explicit decision. Blank is 0. `UNRESOLVED` is 84. Settled is 255 (`ACCEPT` 204, `RECLASSIFY` 32, `NONE` 19). `OBSERVED` 128 and `INFERRED` 127 are separate from the decision. `source_hint` was not copied into `semantic_family`.
 
 The eighteen families were not expanded. One economics row was left `UNRESOLVED` because the fitting region is the inactive candidate `finance-retail`. `brainrot-aura` was not added. `evaluation.enabled` stays false. Rights-blocked rows can carry a semantic decision and still cannot enter `EVAL_RESERVE`. Vendor calls: 0. SELECT-003 was not drafted. Do not train.
+
+
+## ai-native activated — 2026-09-27
+
+The operator authorized `ai-native` as its own evaluation family. It is `taxonomy.active`. `evaluation.enabled` stays false. `production.enabled` stays false. `layout.FAMILIES` is unchanged and still has nine names, including the existing `ai-native` slot. This activation does not settle a row, does not append a ledger, and does not enable evaluation.
+
+A hint of `ai-native` may still be settled as `technology-ai` when the operator writes that family. Writing `semantic_family=ai-native` is now a valid explicit choice. The stored registry class is not promoted to `OBSERVED`.
diff --git a/tests/shadow/test_hyperlexical_eval_settlement.py b/tests/shadow/test_hyperlexical_eval_settlement.py
index d4f70eb4..603bf577 100644
--- a/tests/shadow/test_hyperlexical_eval_settlement.py
+++ b/tests/shadow/test_hyperlexical_eval_settlement.py
@@ -380,8 +380,38 @@ def test_blank_sheet_settles_nothing_and_does_not_fill_decisions(tmp_path):
     assert sheet.read_bytes() == before
 
 
+def test_ai_native_is_taxonomy_active_and_the_head_stays_nine():
+    assert "ai-native" in ACTIVE_FAMILIES
+    assert family_flags("ai-native") == {
+        "taxonomy.active": True,
+        "evaluation.enabled": False,
+        "production.enabled": False,
+    }
+    assert len(PRODUCTION_FAMILIES) == 9
+    assert "ai-native" in PRODUCTION_FAMILIES
+    stream = _stream(
+        "phrase native",
+        hint="ai-native",
+        label=None,
+        label_source="UNLABELLED",
+    )
+    record = _apply(
+        stream,
+        _decision(
+            stream,
+            semantic_family="ai-native",
+            proposed_family="ai-native",
+            lane="B",
+            decision="ACCEPT FAMILY",
+        ),
+    )["records"][0]
+    assert record["semantic_family"] == "ai-native"
+    assert record["source_hint"] == "ai-native"
+    assert record["attest"] == "INFERRED"
+
+
 def test_three_taxonomy_flags_stay_distinct():
-    assert len(ACTIVE_FAMILIES) == 18
+    assert len(ACTIVE_FAMILIES) == 19
     for name in ACTIVE_FAMILIES:
         assert family_flags(name) == {
             "taxonomy.active": True,