diff -r ed0c30b4c082 Lib/json/__init__.py
--- a/Lib/json/__init__.py	Sat May 04 15:16:16 2013 +0300
+++ b/Lib/json/__init__.py	Sun May 05 16:01:07 2013 +0300
@@ -108,6 +108,7 @@
 
 from .decoder import JSONDecoder
 from .encoder import JSONEncoder
+import codecs
 
 _default_encoder = JSONEncoder(
     skipkeys=False,
@@ -243,6 +244,28 @@
 _default_decoder = JSONDecoder(object_hook=None, object_pairs_hook=None)
 
 
+def detect_encoding(b):
+    bstartswith = b.startswith
+    if bstartswith((codecs.BOM_UTF32_BE, codecs.BOM_UTF32_LE)):
+        return 'utf-32'
+    if bstartswith((codecs.BOM_UTF16_BE, codecs.BOM_UTF16_LE)):
+        return 'utf-16'
+    if bstartswith(codecs.BOM_UTF8):
+        return 'utf-8-sig'
+
+    if len(b) >= 4:
+        if not b[0]:
+            return 'utf-16be' if b[1] or b[2] else 'utf-32be'
+        if not b[1]:
+            return 'utf-16le' if b[2] or b[3] else 'utf-32le'
+    elif len(b) == 2:
+        if not b[0]:
+            return 'utf-16be'
+        if not b[1]:
+            return 'utf-16le'
+    return 'utf-8'
+
+
 def load(fp, cls=None, object_hook=None, parse_float=None,
         parse_int=None, parse_constant=None, object_pairs_hook=None, **kw):
     """Deserialize ``fp`` (a ``.read()``-supporting file-like object containing
@@ -310,6 +333,8 @@
     The ``encoding`` argument is ignored and deprecated.
 
     """
+    if isinstance(s, bytes):
+        s = s.decode(detect_encoding(s), 'surrogatepass')
     if (cls is None and object_hook is None and
             parse_int is None and parse_float is None and
             parse_constant is None and object_pairs_hook is None and not kw):
diff -r ed0c30b4c082 Lib/test/json_tests/test_unicode.py
--- a/Lib/test/json_tests/test_unicode.py	Sat May 04 15:16:16 2013 +0300
+++ b/Lib/test/json_tests/test_unicode.py	Sun May 05 16:01:07 2013 +0300
@@ -1,3 +1,4 @@
+import codecs
 from collections import OrderedDict
 from test.json_tests import PyTest, CTest
 
@@ -52,8 +53,18 @@
         self.assertRaises(TypeError, self.dumps, [b"hi"])
 
     def test_bytes_decode(self):
-        self.assertRaises(TypeError, self.loads, b'"hi"')
-        self.assertRaises(TypeError, self.loads, b'["hi"]')
+        for encoding, bom in [
+                ('utf-8', codecs.BOM_UTF8),
+                ('utf-16be', codecs.BOM_UTF16_BE),
+                ('utf-16le', codecs.BOM_UTF16_LE),
+                ('utf-32be', codecs.BOM_UTF32_BE),
+                ('utf-32le', codecs.BOM_UTF32_LE),
+            ]:
+            data = ["a\xb5\u20ac\U0001d120"]
+            encoded = self.dumps(data).encode(encoding)
+            self.assertEqual(self.loads(bom + encoded), data)
+            self.assertEqual(self.loads(encoded), data)
+        self.assertRaises(UnicodeDecodeError, self.loads, b'["\x80"]')
 
 
     def test_object_pairs_hook_with_unicode(self):