diff --git a/CHANGES.md b/CHANGES.md
index 1194fef1..673ddfc0 100644
--- a/CHANGES.md
+++ b/CHANGES.md
@@ -15,6 +15,12 @@
   ``"prefix"`` was checked first.
   https://github.com/joblib/threadpoolctl/pull/278
 
+- Fixed `limits="sequential_blas_under_openmp"`, which left all BLAS libraries
+  unlimited as soon as OpenBLAS was loaded, whatever its threading layer, or a
+  library used the OpenMP threading layer (e.g. BLIS). The limit is now only
+  skipped when OpenBLAS uses the OpenMP threading layer, as documented.
+  https://github.com/joblib/threadpoolctl/pull/283
+
 
 3.7.0 (2026-09-15)
 ==================
diff --git a/tests/test_threadpoolctl.py b/tests/test_threadpoolctl.py
index fc2c491f..922d6e76 100644
--- a/tests/test_threadpoolctl.py
+++ b/tests/test_threadpoolctl.py
@@ -299,9 +299,11 @@ def test_get_params_for_sequential_blas_under_openmp():
 
     params = controller._get_params_for_sequential_blas_under_openmp()
 
-    if controller.select(
-        internal_api="openblas", threading_layer="openmp"
-    ).lib_controllers:
+    if any(
+        lib_info["internal_api"] == "openblas"
+        and lib_info.get("threading_layer") == "openmp"
+        for lib_info in original_info
+    ):
         assert params["limits"] is None
         assert params["user_api"] is None
 
diff --git a/threadpoolctl.py b/threadpoolctl.py
index b883cae0..6f2da642 100644
--- a/threadpoolctl.py
+++ b/threadpoolctl.py
@@ -1054,9 +1054,12 @@ def _get_params_for_sequential_blas_under_openmp(self):
         This function takes into account the unexpected behavior of OpenBLAS with the
         OpenMP threading layer.
         """
-        if self.select(
-            internal_api="openblas", threading_layer="openmp"
-        ).lib_controllers:
+        # select() matches any of its kwargs: chain two calls to require both.
+        if (
+            self.select(internal_api="openblas")
+            .select(threading_layer="openmp")
+            .lib_controllers
+        ):
             # Don't change anything: OpenBLAS is already sequential by default when it
             # is called from an OpenMP parallel region, and limiting it would also
             # limit the OpenMP parallel regions of the calling thread.