diff --git a/src/backend/InvenTree/importer/models.py b/src/backend/InvenTree/importer/models.py
index 8c62ac5d3e..dcd4601c4f 100644
--- a/src/backend/InvenTree/importer/models.py
+++ b/src/backend/InvenTree/importer/models.py
@@ -237,7 +237,14 @@ class DataImportSession(models.Model):
 
             # Extract a "default" value for the field, if one exists
             # Skip if one has already been provided by the user
-            if field not in self.field_defaults and 'default' in field_def:
+            # Skip entirely when updating existing records - a model-level "creation"
+            # default (e.g. quantity=1) must not silently overwrite existing data
+            # for fields the user hasn't chosen to map or default
+            if (
+                not self.update_records
+                and field not in self.field_defaults
+                and 'default' in field_def
+            ):
                 self.field_defaults[field] = field_def['default']
 
             # Generate a list of possible column names for this field
@@ -472,11 +479,14 @@ class DataImportSession(models.Model):
         required = {}
 
         for field, info in fields.items():
-            if info.get('required', False):
-                required[field] = info
-
-            elif self.update_records and field == self.ID_FIELD_LABEL:
-                # If we are updating records, the ID field is required
+            if self.update_records:
+                # When updating existing records, only the ID field is required -
+                # all other fields fall back to the existing value on the record
+                # being updated, so the serializer's usual "required" flag
+                # (which applies to *creating* a new instance) does not apply here
+                if field == self.ID_FIELD_LABEL:
+                    required[field] = info
+            elif info.get('required', False):
                 required[field] = info
 
         return required
@@ -933,6 +943,11 @@ class DataImportRow(models.Model):
             return serializer_class(
                 instance=instance,
                 data=self.serializer_data(),
+                # When updating an existing instance, treat this as a partial
+                # update - fields not present in the imported data should fall
+                # back to the existing value on the record, not fail validation
+                # or be overwritten with a blank/default value
+                partial=instance is not None,
                 context={'request': request},
             )
 
diff --git a/src/backend/InvenTree/importer/tests.py b/src/backend/InvenTree/importer/tests.py
index f0dfde284d..2b5a06a007 100644
--- a/src/backend/InvenTree/importer/tests.py
+++ b/src/backend/InvenTree/importer/tests.py
@@ -194,6 +194,68 @@ class ImporterTest(ImporterMixin, InvenTreeTestCase):
         with self.assertRaises(DjangoValidationError):
             session.full_clean()
 
+    def test_update_existing_records(self):
+        """Test updating existing records via import, providing only the ID field.
+
+        Regression test for GH #12499: when updating existing records, only the
+        primary key should be required. Fields which are not mapped to a column
+        (and have no explicit override/default) must:
+        - not be flagged as "required" and block the mapping wizard
+        - not be auto-populated with a model-level "creation" default
+        - not overwrite the existing value on the target record
+        """
+        from part.models import Part, PartCategory
+        from stock.models import StockItem
+
+        category = PartCategory.objects.create(
+            name='Update Test Category', description='Test category'
+        )
+        part = Part.objects.create(
+            category=category, name='Update Test Part', description='Test part'
+        )
+
+        item = StockItem.objects.create(part=part, quantity=42, batch='Original batch')
+
+        csv_content = f'ID,batch\n{item.pk},Updated batch\n'
+        data_file = ContentFile(csv_content, 'stock_update.csv')
+
+        session = DataImportSession.objects.create(
+            data_file=data_file, model_type='stockitem', update_records=True
+        )
+
+        # Only the 'id' field should be required - 'quantity' and 'part' already
+        # exist on the target record, and must not be forced
+        required = session.required_fields()
+        self.assertIn('id', required)
+        self.assertNotIn('quantity', required)
+        self.assertNotIn('part', required)
+
+        # No model-level "creation" default should have been auto-populated for
+        # 'quantity' - doing so would silently overwrite the existing value
+        self.assertNotIn('quantity', session.field_defaults or {})
+
+        # Mapping is valid, even though 'quantity' and 'part' are unmapped
+        session.accept_mapping()
+
+        session.refresh_from_db()
+        self.assertEqual(session.rows.count(), 1)
+
+        row = session.rows.first()
+        self.assertTrue(row.valid)
+
+        # Unmapped fields must not appear in the extracted row data at all
+        self.assertNotIn('quantity', row.data)
+        self.assertNotIn('part', row.data)
+
+        self.assertTrue(row.validate(commit=True))
+        self.assertTrue(row.complete)
+
+        # Existing values for unmapped fields must be preserved
+        item.refresh_from_db()
+        self.assertEqual(item.quantity, 42)
+        self.assertEqual(item.part, part)
+        self.assertEqual(item.batch, 'Updated batch')
+
     def test_lookup_field_ambiguous_match(self):
         """Test the behavior of lookup_related_field for ambiguous and pinned matches."""
         from django.core.exceptions import ValidationError as DjangoValidationError