diff --git a/core/wren/src/wren/policy.py b/core/wren/src/wren/policy.py
index 9ded7e3a5e..54fff54bc0 100644
--- a/core/wren/src/wren/policy.py
+++ b/core/wren/src/wren/policy.py
@@ -156,9 +156,10 @@ def resolve_model_name(
     falls back to a case-insensitive scan. Returns ``None`` if no model
     matches.
     """
-    model_set = (
-        model_names if isinstance(model_names, (set, frozenset)) else set(model_names)
-    )
+    # Drop non-str names so a corrupt models collection cannot crash .lower().
+    if not isinstance(name, str) or not name:
+        return None
+    model_set = {n for n in model_names if isinstance(n, str) and n}
     if name in model_set:
         return name
     if quoted:
diff --git a/core/wren/tests/unit/test_resolve_model_name_nonstr.py b/core/wren/tests/unit/test_resolve_model_name_nonstr.py
new file mode 100644
index 0000000000..68daab26be
--- /dev/null
+++ b/core/wren/tests/unit/test_resolve_model_name_nonstr.py
@@ -0,0 +1,25 @@
+"""resolve_model_name must ignore non-str names without AttributeError."""
+
+from __future__ import annotations
+
+import pytest
+
+from wren.policy import resolve_model_name
+
+pytestmark = pytest.mark.unit
+
+
+def test_resolve_filters_non_str_candidates() -> None:
+    names = {"Orders", None, 12, "", "customers"}  # type: ignore[list-item]
+    assert resolve_model_name("orders", quoted=False, model_names=names) == "Orders"
+    assert resolve_model_name("missing", quoted=False, model_names=names) is None
+
+
+def test_resolve_rejects_non_str_lookup() -> None:
+    assert resolve_model_name(None, quoted=False, model_names={"a"}) is None  # type: ignore[arg-type]
+    assert resolve_model_name(1, quoted=True, model_names={"1"}) is None  # type: ignore[arg-type]
+
+
+def test_resolve_quoted_exact() -> None:
+    assert resolve_model_name("Orders", quoted=True, model_names={"Orders"}) == "Orders"
+    assert resolve_model_name("orders", quoted=True, model_names={"Orders"}) is None