diff --git a/tests/rules/test_quoted_strings.py b/tests/rules/test_quoted_strings.py
index 1bcb6f8f..31b392a9 100644
--- a/tests/rules/test_quoted_strings.py
+++ b/tests/rules/test_quoted_strings.py
@@ -14,6 +14,7 @@
 # along with this program.  If not, see <http://www.gnu.org/licenses/>.
 
 from tests.common import RuleTestCase
+import yaml.reader
 
 from yamllint import config
 
@@ -475,6 +476,30 @@ def test_only_when_needed_extras(self):
                    '- "foo bar"\n',
                    conf, problem1=(3, 3), problem2=(7, 3), problem3=(11, 3))
 
+    def test_only_when_needed_special_characters(self):
+        conf = 'quoted-strings: {required: only-when-needed}\n'
+        self.check('---\n'
+                   'k1: "\\u001b"\n',
+                   conf)
+        self.check('---\n'
+                   "k1: '\\u001b'\n",
+                   conf, problem=(2, 5))
+        self.check('---\n'
+                   'k1: \\u001b\n',
+                   conf)
+        self.assertRaises(yaml.reader.ReaderError, self.check,
+                          '---\n'
+                          'k1: "\u001b"\n',
+                          conf)
+        self.assertRaises(yaml.reader.ReaderError, self.check,
+                          '---\n'
+                          "k1: '\u001b'\n",
+                          conf)
+        self.assertRaises(yaml.reader.ReaderError, self.check,
+                          '---\n'
+                          'k1: \u001b\n',
+                          conf)
+
     def test_octal_values(self):
         conf = 'quoted-strings: {required: true}\n'
 
diff --git a/yamllint/rules/quoted_strings.py b/yamllint/rules/quoted_strings.py
index b38f6edf..a013dc6b 100644
--- a/yamllint/rules/quoted_strings.py
+++ b/yamllint/rules/quoted_strings.py
@@ -204,11 +204,19 @@ def _quote_match(quote_type, token_style):
             (quote_type == 'double' and token_style == '"'))
 
 
-def _quotes_are_needed(string, is_inside_a_flow):
+def _quotes_are_needed(string, style, is_inside_a_flow):
     # Quotes needed on strings containing flow tokens
     if is_inside_a_flow and set(string) & {',', '[', ']', '{', '}'}:
         return True
 
+    if style == '"':
+        try:
+            yaml.reader.Reader('').check_printable('key: ' + string)
+        except yaml.reader.ReaderError:
+            # Special characters in a double-quoted string
+            # are assumed to have been backslash-escaped
+            return True
+
     loader = yaml.BaseLoader('key: ' + string)
     # Remove the 5 first tokens corresponding to 'key: ' (StreamStartToken,
     # BlockMappingStartToken, KeyToken, ScalarToken(value=key), ValueToken)
@@ -299,6 +307,7 @@ def check(conf, token, prev, next, nextnext, context):
         # Quotes are not strictly needed here
         if (token.style and tag == DEFAULT_SCALAR_TAG and token.value and
                 not _quotes_are_needed(token.value,
+                                       token.style,
                                        context['flow_nest_count'] > 0)):
             is_extra_required = any(re.search(r, token.value)
                                     for r in conf['extra-required'])