diff --git a/autofit/mapper/prior/log_uniform.py b/autofit/mapper/prior/log_uniform.py
index 4958bcd5e..7d0725468 100644
--- a/autofit/mapper/prior/log_uniform.py
+++ b/autofit/mapper/prior/log_uniform.py
@@ -123,10 +123,21 @@ def log_prior_from_value(self, value, xp=np):
         used by ``UniformPrior.log_prior_from_value`` which drops ``-log(b - a)``
         to return ``0.0``.
 
-        Out-of-support (``value`` outside ``[lower_limit, upper_limit]``) returns
-        ``-inf`` on the JAX path; the NumPy path returns ``-log(value)`` without
-        bounds-checking to mirror the existing ``UniformPrior`` NumPy semantics
-        which trust the search to stay inside bounds.
+        Non-positive ``value`` (``value <= 0``) returns ``-inf``. Emcee's stretch
+        move proposes physical values that can leave the support and go
+        non-positive; ``-log`` of a non-positive value is ``NaN``, which propagates
+        into the summed figure-of-merit and crashes the search with
+        ``ValueError: Probability function returned NaN``. Returning ``-inf``
+        (zero density -> rejected move) keeps the figure-of-merit finite. The
+        "double where" pattern (a safe surrogate inside the ``log``) ensures no
+        ``log`` of a non-positive value is evaluated, avoiding NumPy
+        ``RuntimeWarning``s.
+
+        The NumPy path is otherwise unnormalised and unbounded: for any positive
+        ``value`` it returns ``-log(value)`` regardless of ``[lower_limit,
+        upper_limit]`` (dropping the normalisation constant, matching
+        ``UniformPrior``'s convention of returning ``0.0``). The JAX path
+        additionally returns ``-inf`` outside ``[lower_limit, upper_limit]``.
 
         Parameters
         ----------
@@ -137,7 +148,8 @@ def log_prior_from_value(self, value, xp=np):
             Array-module to dispatch on (``numpy`` or ``jax.numpy``). Default ``numpy``.
         """
         if xp is np:
-            return -np.log(value)
+            positive = value > 0.0
+            return xp.where(positive, -xp.log(xp.where(positive, value, 1.0)), -xp.inf)
         in_bounds = (value >= self.lower_limit) & (value <= self.upper_limit)
         return xp.where(in_bounds, -xp.log(value), -xp.inf)
 
diff --git a/test_autofit/mapper/prior/test_prior.py b/test_autofit/mapper/prior/test_prior.py
index a2b4895a9..730a95073 100644
--- a/test_autofit/mapper/prior/test_prior.py
+++ b/test_autofit/mapper/prior/test_prior.py
@@ -1,4 +1,5 @@
 import math
+import warnings
 
 import numpy as np
 import pytest
@@ -216,6 +217,30 @@ def test__log_prior_from_value(self):
             -np.log(4.0), 1.0e-12
         )
 
+    def test__log_prior_from_value__non_positive_returns_neg_inf(self):
+        # Regression (PyAutoHeart #27 / release run 28784914443): Emcee's stretch
+        # move proposes physical values that can go non-positive for a LogUniform
+        # parameter. `-log(value)` of a non-positive value is NaN, which propagated
+        # into the summed figure-of-merit and crashed the search with
+        # "ValueError: Probability function returned NaN". Non-positive values must
+        # return -inf (zero density -> the move is rejected), and evaluating the
+        # log-prior must not emit a NumPy "invalid value in log" RuntimeWarning.
+        log_uniform = af.LogUniformPrior(lower_limit=1e-3, upper_limit=1e3)
+
+        with warnings.catch_warnings():
+            warnings.simplefilter("error", category=RuntimeWarning)
+            assert log_uniform.log_prior_from_value(value=-1.0) == float("-inf")
+            assert log_uniform.log_prior_from_value(value=0.0) == float("-inf")
+
+        # Positive values are unchanged: the NumPy path stays unnormalised and
+        # unbounded, returning -log(value) regardless of (lower_limit, upper_limit).
+        assert log_uniform.log_prior_from_value(value=10.0) == pytest.approx(
+            -np.log(10.0), 1.0e-12
+        )
+        assert log_uniform.log_prior_from_value(value=1.0e4) == pytest.approx(
+            -np.log(1.0e4), 1.0e-12
+        )
+
     def test__lower_limit_zero_or_below_raises_error(self):
         with pytest.raises(exc.PriorException):
             af.LogUniformPrior(lower_limit=-1.0, upper_limit=1.0)