diff --git a/src/pyrecest/filters/dirichlet_process_birth_tracker.py b/src/pyrecest/filters/dirichlet_process_birth_tracker.py
index 0a812478c..e10755c0c 100644
--- a/src/pyrecest/filters/dirichlet_process_birth_tracker.py
+++ b/src/pyrecest/filters/dirichlet_process_birth_tracker.py
@@ -238,29 +238,44 @@ def _create_birth_component_from_measurement(
             birth_mean,
             birth_covariance,
         )
-        self.last_birth_diagnostics.append(decision)
         if decision["action"] == "clutter":
+            self.last_birth_diagnostics.append(decision)
             return None
 
+        pruning_threshold = self._normalize_birth_atom_pruning_threshold()
+        maximum_number_of_birth_atoms = self._normalize_maximum_number_of_birth_atoms()
+
+        atom_index = decision["atom_index"]
         if decision["action"] == "new_atom":
             atom = DirichletProcessBirthAtom(birth_mean, birth_covariance)
-            self.birth_atoms.append(atom)
-            decision["atom_index"] = len(self.birth_atoms) - 1
+            atom_index = len(self.birth_atoms)
+            decision["atom_index"] = atom_index
         else:
-            atom = self.birth_atoms[decision["atom_index"]]
+            atom = self.birth_atoms[atom_index].copy()
             atom.update_from_measurement(
                 measurement,
                 measurement_matrix,
                 measurement_covariance,
             )
 
-        self._prune_and_cap_birth_atoms()
-        return BernoulliComponent(
+        birth_component = BernoulliComponent(
             decision["existence_probability"],
             GaussianDistribution(atom.mean, atom.covariance),
             label=label,
         )
 
+        if decision["action"] == "new_atom":
+            self.birth_atoms.append(atom)
+        else:
+            self.birth_atoms[atom_index] = atom
+
+        self._prune_and_cap_birth_atoms(
+            pruning_threshold=pruning_threshold,
+            maximum_number_of_birth_atoms=maximum_number_of_birth_atoms,
+        )
+        self.last_birth_diagnostics.append(decision)
+        return birth_component
+
     def _resolve_birth_covariance(
         self,
         measurement,
diff --git a/tests/filters/test_dirichlet_process_birth_atomicity.py b/tests/filters/test_dirichlet_process_birth_atomicity.py
new file mode 100644
index 000000000..c8e8c975e
--- /dev/null
+++ b/tests/filters/test_dirichlet_process_birth_atomicity.py
@@ -0,0 +1,106 @@
+import numpy as np
+import pyrecest.backend
+import pytest
+from pyrecest.filters.dirichlet_process_birth_tracker import (
+    DirichletProcessBirthMultiBernoulliTracker,
+)
+
+pytestmark = pytest.mark.skipif(
+    pyrecest.backend.__backend_name__ != "numpy",
+    reason="DP birth multi-Bernoulli tracker is NumPy-only.",
+)
+
+
+def _tracker(**overrides):
+    birth_atoms = overrides.pop("birth_atoms", None)
+    tracker_param = {
+        "birth_covariance": np.diag([1.0, 1.0, 4.0, 4.0]),
+        "birth_existence_probability": 0.8,
+        "clutter_intensity": 1e-6,
+        "dp_concentration": 0.05,
+        "dp_birth_threshold": 1.0,
+        "measurement_to_state_matrix": np.array(
+            [
+                [1.0, 0.0],
+                [0.0, 1.0],
+                [0.0, 0.0],
+                [0.0, 0.0],
+            ]
+        ),
+    }
+    tracker_param.update(overrides)
+    measurement_matrix = np.array(
+        [
+            [1.0, 0.0, 0.0, 0.0],
+            [0.0, 1.0, 0.0, 0.0],
+        ]
+    )
+    measurement_covariance = np.eye(2) * 0.2
+    return (
+        DirichletProcessBirthMultiBernoulliTracker(
+            tracker_param=tracker_param,
+            birth_atoms=birth_atoms,
+        ),
+        measurement_matrix,
+        measurement_covariance,
+    )
+
+
+def _assert_atom_unchanged(atom, reference):
+    np.testing.assert_allclose(atom.mean, reference.mean)
+    np.testing.assert_allclose(atom.covariance, reference.covariance)
+    assert atom.count == reference.count
+
+
+def test_new_birth_rejects_invalid_atom_cap_without_partial_state():
+    tracker, measurement_matrix, measurement_covariance = _tracker(
+        maximum_number_of_birth_atoms=1.5
+    )
+
+    with pytest.raises(ValueError, match="maximum_number_of_birth_atoms"):
+        tracker._create_birth_component_from_measurement(
+            np.array([2.0, 3.0]),
+            measurement_matrix,
+            measurement_covariance,
+        )
+
+    assert tracker.birth_atoms == []
+    assert tracker.last_birth_diagnostics == []
+
+
+def test_existing_birth_rejects_invalid_pruning_without_mutating_atom():
+    tracker, measurement_matrix, measurement_covariance = _tracker(
+        birth_atoms=[(np.zeros(4), np.eye(4), 2.0)],
+        dp_birth_atom_pruning_threshold=np.nan,
+    )
+    atom_before = tracker.get_birth_atoms()[0]
+
+    with pytest.raises(ValueError, match="dp_birth_atom_pruning_threshold"):
+        tracker._create_birth_component_from_measurement(
+            np.array([0.1, -0.1]),
+            measurement_matrix,
+            measurement_covariance,
+        )
+
+    assert len(tracker.birth_atoms) == 1
+    _assert_atom_unchanged(tracker.birth_atoms[0], atom_before)
+    assert tracker.last_birth_diagnostics == []
+
+
+def test_failed_birth_component_construction_does_not_mutate_existing_atom():
+    tracker, measurement_matrix, measurement_covariance = _tracker(
+        birth_atoms=[(np.zeros(4), np.eye(4), 2.0)],
+        birth_existence_probability=np.nan,
+    )
+    atom_before = tracker.get_birth_atoms()[0]
+
+    with pytest.raises(ValueError, match="existence_probability"):
+        tracker._create_birth_component_from_measurement(
+            np.array([0.1, -0.1]),
+            measurement_matrix,
+            measurement_covariance,
+        )
+
+    assert len(tracker.birth_atoms) == 1
+    _assert_atom_unchanged(tracker.birth_atoms[0], atom_before)
+    assert tracker.last_birth_diagnostics == []