diff --git a/autofit/non_linear/search/nest/nautilus/search.py b/autofit/non_linear/search/nest/nautilus/search.py
index 4ae265cc6..2cb9dce56 100644
--- a/autofit/non_linear/search/nest/nautilus/search.py
+++ b/autofit/non_linear/search/nest/nautilus/search.py
@@ -2,6 +2,7 @@
 import logging
 import os
 import sys
+from contextlib import nullcontext
 from pathlib import Path
 from typing import Dict, Optional, Tuple
 
@@ -327,7 +328,18 @@ def fit_multiprocessing(self, fitness, model, analysis):
         # A pool object is passed rather than pool=<int> so the pool uses the
         # "fork" start method (see autofit.non_linear.parallel.fork_context) —
         # nautilus builds its internal pools from the default context.
-        with fork_context().Pool(self.number_of_cores) as pool:
+        #
+        # For a single core no pool may be created at all: nautilus treats
+        # pool=None (and the int 1) as fully serial, whereas a Pool(1) object
+        # forces every likelihood call into a forked worker — which deadlocks
+        # in XLA compilation when the likelihood touches JAX, since a forked
+        # child of a JAX-initialized parent cannot compile.
+        if self.number_of_cores <= 1:
+            pool_context = nullcontext(None)
+        else:
+            pool_context = fork_context().Pool(self.number_of_cores)
+
+        with pool_context as pool:
             search_internal = self.sampler_cls(
                 prior=PriorVectorized(model=model),
                 likelihood=fitness.call_wrap,
diff --git a/test_autofit/non_linear/search/nest/test_nautilus.py b/test_autofit/non_linear/search/nest/test_nautilus.py
index 5ac5ff357..ca217876b 100644
--- a/test_autofit/non_linear/search/nest/test_nautilus.py
+++ b/test_autofit/non_linear/search/nest/test_nautilus.py
@@ -1,3 +1,4 @@
+import numpy as np
 import pytest
 
 import autofit as af
@@ -51,3 +52,36 @@ def test__test_mode():
     search.apply_test_mode()
 
     assert search.n_like_max == 1
+
+
+def test__single_core_builds_no_pool(monkeypatch):
+    """
+    number_of_cores=1 must not construct a multiprocessing pool: nautilus
+    treats pool=None as fully serial, whereas a Pool(1) object forces every
+    likelihood call into a forked worker — which deadlocks in XLA compilation
+    when the likelihood touches JAX (#1442).
+    """
+    from autofit.non_linear.search.nest.nautilus import search as nautilus_search
+
+    def no_fork_context():
+        raise AssertionError(
+            "fork_context must not be used when number_of_cores == 1"
+        )
+
+    monkeypatch.setattr(nautilus_search, "fork_context", no_fork_context)
+    monkeypatch.setenv("PYAUTO_TEST_MODE", "1")
+
+    model = af.Model(af.ex.Gaussian)
+    analysis = af.ex.Analysis(
+        data=np.full(100, 5.0),
+        noise_map=np.full(100, 1.0),
+    )
+
+    search = af.Nautilus(
+        name="nautilus_single_core",
+        unique_tag="single_core_no_pool_test",
+        n_live=10,
+        number_of_cores=1,
+    )
+
+    search.fit(model=model, analysis=analysis)