diff --git a/CHANGELOG.md b/CHANGELOG.md
index 64cd471d..4b7a65b9 100644
--- a/CHANGELOG.md
+++ b/CHANGELOG.md
@@ -17,6 +17,7 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 * Anthropic-backed providers (`ChatAnthropic()`, `ChatPosit()`, `ChatBedrock()`, etc.) no longer fail with `Invalid signature in thinking block` when the conversation history contains reasoning from a non-Claude model (e.g., after switching `chat.model` across families); such thinking is now replayed as plain text so the model can still see it.
 * `content_image_file()` no longer fails with `ValueError: unknown file extension` when `resize` uses the `!` (ignore aspect ratio) flag on an image larger than the requested box, e.g. `resize="200x200!"`. (#433)
 * `params(top_k=)` is no longer sent as `top_logprobs` for OpenAI-based providers (the two are unrelated; OpenAI has no `top_k` sampling parameter). `top_k` is now dropped with the standard unsupported-parameter warning. (#412)
+* `Chat.export()` and `interpolate_file()` now always read and write files as UTF-8. Previously they relied on the platform's default locale encoding, so on a non-UTF-8 locale `export()` could raise `UnicodeEncodeError` and `interpolate_file()` could silently return a corrupted (mojibake) prompt instead of the file's contents. (#437)
 
 
 ## [0.23.0] - 2026-09-04
diff --git a/chatlas/_chat.py b/chatlas/_chat.py
index 36e592ab..82d742bd 100644
--- a/chatlas/_chat.py
+++ b/chatlas/_chat.py
@@ -2584,7 +2584,7 @@ def export(
         if is_html:
             contents = self._html_template(contents)
 
-        with open(filename, "w") as f:
+        with open(filename, "w", encoding="utf-8") as f:
             f.write(contents)
 
         return filename
diff --git a/chatlas/_interpolate.py b/chatlas/_interpolate.py
index 7047ab86..5bfb001f 100644
--- a/chatlas/_interpolate.py
+++ b/chatlas/_interpolate.py
@@ -109,7 +109,7 @@ def interpolate_file(
         variables = _infer_variables(frame)
         del frame
 
-    with open(path, "r") as file:
+    with open(path, "r", encoding="utf-8") as file:
         return interpolate(
             file.read(),
             variables=variables,
diff --git a/tests/test_chat.py b/tests/test_chat.py
index e84aec28..07ecf3a2 100644
--- a/tests/test_chat.py
+++ b/tests/test_chat.py
@@ -161,6 +161,39 @@ def test_basic_export(snapshot):
             assert snapshot == f.read()
 
 
+def test_export_writes_utf8(monkeypatch):
+    # `Chat.export()` must write the transcript as UTF-8 regardless of the
+    # platform's default locale encoding, otherwise non-ASCII turn content
+    # raises UnicodeEncodeError instead of exporting successfully.
+    text = "Wie heißt die Hauptstadt? 中文测试 café"
+    chat = ChatOpenAI(
+        system_prompt="You're a helpful assistant that returns very minimal output",
+    )
+    chat.set_turns(
+        [
+            UserTurn(text),
+            AssistantTurn(text, tokens=(15, 5, 0)),
+        ]
+    )
+
+    real_open = open
+
+    def open_using_locale_default(file, mode="r", *args, encoding=None, **kwargs):
+        # Simulate a non-UTF-8 preferred locale encoding by falling back to
+        # latin-1 whenever the caller doesn't pass an explicit encoding.
+        if "b" not in mode and encoding is None:
+            encoding = "latin-1"
+        return real_open(file, mode, *args, encoding=encoding, **kwargs)
+
+    monkeypatch.setattr("chatlas._chat.open", open_using_locale_default, raising=False)
+
+    with tempfile.TemporaryDirectory() as tmpdir:
+        tmpfile = tmpdir + "/chat.md"
+        chat.export(tmpfile)
+        with open(tmpfile, "r", encoding="utf-8") as f:
+            assert text in f.read()
+
+
 @pytest.mark.vcr
 def test_chat_structured():
     chat = ChatOpenAI(model="gpt-5.4")
diff --git a/tests/test_interpolate.py b/tests/test_interpolate.py
index 9b2f71a1..70be0b8d 100644
--- a/tests/test_interpolate.py
+++ b/tests/test_interpolate.py
@@ -18,3 +18,28 @@ def test_interpolate_file(tmp_path):
 
         x = 1  # noqa
         assert interpolate_file(path) == "1"
+
+
+def test_interpolate_file_reads_as_utf8(tmp_path, monkeypatch):
+    # `interpolate_file()` must read the prompt file as UTF-8 regardless of
+    # the platform's default locale encoding, otherwise non-ASCII prompt
+    # content is silently corrupted (mojibake) instead of raising.
+    text = "Wie heißt die Hauptstadt? 中文测试 café {{ name }}"
+    path = tmp_path / "prompt.txt"
+    path.write_text(text, encoding="utf-8")
+
+    real_open = open
+
+    def open_using_locale_default(file, mode="r", *args, encoding=None, **kwargs):
+        # Simulate a non-UTF-8 preferred locale encoding by falling back to
+        # latin-1 whenever the caller doesn't pass an explicit encoding.
+        if "b" not in mode and encoding is None:
+            encoding = "latin-1"
+        return real_open(file, mode, *args, encoding=encoding, **kwargs)
+
+    monkeypatch.setattr(
+        "chatlas._interpolate.open", open_using_locale_default, raising=False
+    )
+
+    result = interpolate_file(path, variables={"name": "x"})
+    assert "Wie heißt die Hauptstadt? 中文测试 café" in result