diff --git a/monai/transforms/intensity/array.py b/monai/transforms/intensity/array.py
index ec0a579c01..ff5462e25e 100644
--- a/monai/transforms/intensity/array.py
+++ b/monai/transforms/intensity/array.py
@@ -254,7 +254,10 @@ def __call__(self, img: NdarrayOrTensor, offset: float | None = None) -> Ndarray
 
         img = convert_to_tensor(img, track_meta=get_track_meta())
         offset = self.offset if offset is None else offset
-        out = img + offset
+        # Promote integer inputs before addition when safe=True. Otherwise integer
+        # offsets can overflow in the input dtype before safe conversion clamps.
+        computation_img = img.to(torch.float64) if self.safe and not img.is_floating_point() else img
+        out = computation_img + offset
         out, *_ = convert_data_type(data=out, dtype=img.dtype, safe=self.safe)
 
         return out
diff --git a/tests/transforms/test_shift_intensity.py b/tests/transforms/test_shift_intensity.py
index 70a6c7bf79..4346f9adc9 100644
--- a/tests/transforms/test_shift_intensity.py
+++ b/tests/transforms/test_shift_intensity.py
@@ -14,12 +14,27 @@
 import unittest
 
 import numpy as np
+import torch
 
 from monai.transforms import ShiftIntensity
 from tests.test_utils import NumpyImageTestCase2D
 
 
 class TestShiftIntensity(NumpyImageTestCase2D):
+    def test_safe_integer_offset_clamps_unsigned_image(self):
+        image = torch.tensor([0, 100, 255], dtype=torch.uint8)
+        for offset, expected in ((1, [1, 101, 255]), (-1, [0, 99, 254]), (1.0, [1, 101, 255]), (-1.0, [0, 99, 254])):
+            with self.subTest(offset=offset):
+                result = ShiftIntensity(offset=offset, safe=True)(image)
+                torch.testing.assert_close(result, torch.tensor(expected, dtype=torch.uint8))
+
+    def test_safe_integer_offset_clamps_signed_image(self):
+        image = torch.tensor([-128, 0, 127], dtype=torch.int8)
+        result = ShiftIntensity(offset=1, safe=True)(image)
+        torch.testing.assert_close(result, torch.tensor([-127, 1, 127], dtype=torch.int8))
+        result = ShiftIntensity(offset=-1, safe=True)(image)
+        torch.testing.assert_close(result, torch.tensor([-128, -1, 126], dtype=torch.int8))
+
     def test_value(self):
         shifter = ShiftIntensity(offset=1.0)
         result = shifter(self.imt)