diff --git a/backend/tests/test_rc6_supply_chain.py b/backend/tests/test_rc6_supply_chain.py
index a0d27e49..619b2572 100644
--- a/backend/tests/test_rc6_supply_chain.py
+++ b/backend/tests/test_rc6_supply_chain.py
@@ -105,7 +105,7 @@ def test_release_image_uses_locked_non_ai_dependencies_and_npm_ci() -> None:
     assert "pip install --no-cache-dir --require-hashes -r requirements-runtime.lock" in dockerfile
     assert "ARG GEOINTEL_ULTRALYTICS_VERSION=8.4.99" in dockerfile
     assert '"ultralytics==$GEOINTEL_ULTRALYTICS_VERSION"' in dockerfile
-    assert "ARG GEOINTEL_SETUPTOOLS_VERSION=83.0.0" in dockerfile
+    assert "ARG GEOINTEL_SETUPTOOLS_VERSION=81.0.0" in dockerfile
     assert "ARG GEOINTEL_WHEEL_VERSION=0.47.0" in dockerfile
     assert "COPY deploy/unraid/gosu-setpriv /usr/local/bin/gosu" in dockerfile
     assert "&& pip check" in dockerfile
diff --git a/deploy/unraid/Dockerfile.all-in-one b/deploy/unraid/Dockerfile.all-in-one
index a428be8f..43a31706 100644
--- a/deploy/unraid/Dockerfile.all-in-one
+++ b/deploy/unraid/Dockerfile.all-in-one
@@ -13,7 +13,7 @@ ARG GEOINTEL_TORCH_INDEX_URL=https://download.pytorch.org/whl/cu128
 ARG GEOINTEL_TORCH_VERSION=2.11.0
 ARG GEOINTEL_TORCHVISION_VERSION=0.26.0
 ARG GEOINTEL_ULTRALYTICS_VERSION=8.4.99
-ARG GEOINTEL_SETUPTOOLS_VERSION=83.0.0
+ARG GEOINTEL_SETUPTOOLS_VERSION=81.0.0
 ARG GEOINTEL_WHEEL_VERSION=0.47.0
 
 ENV GEOINTEL_ENV=production \
diff --git a/docs/CODEX_EXECUTION_LOG.md b/docs/CODEX_EXECUTION_LOG.md
index bdf734c8..3f7fcdd9 100644
--- a/docs/CODEX_EXECUTION_LOG.md
+++ b/docs/CODEX_EXECUTION_LOG.md
@@ -12231,3 +12231,4 @@ Open:
 - TypeScript- en Vite-productiebuild geslaagd.
 - Gerichte frontend-, backend-, browser- en productieverificatie volgen hieronder na de releasegate.
 - De eerste productieacceptatie vond een echte CUDA-compatibiliteitsregressie: `torch 2.13.0+cu130` zag de RTX 4080 SUPER maar weigerde initialisatie op de Tower-driver met CUDA 12.9-capaciteit. De AI-image is daarom teruggebracht naar de expliciet gepinde, drivercompatibele combinatie `torch 2.11.0` / `torchvision 0.26.0` via de officiële CUDA 12.8-index; promotie vereist opnieuw live `torch.cuda.is_available()` en modelpreflight.
+- De CUDA 12.8 PyTorch-wheel vereist `setuptools<82`; de imagepin is daarom samenhangend teruggebracht van 83.0.0 naar 81.0.0 en blijft door `pip check` bewaakt.