diff --git a/cpanfile b/cpanfile
index 3062244eb..a65bc87df 100644
--- a/cpanfile
+++ b/cpanfile
@@ -192,7 +192,7 @@ recommends 'Net::SMTP';
 # Note: Perl 5.10.1 bundles 1.03 (per Unicode 5.1.0).
 recommends 'Unicode::Normalize', '>= 1.03';
 
-recommends 'Unicode::UTF8', '>= 0.58';
+recommends 'Unicode::UTF8', '>= 0.60';
 
 ### Features
 ##
@@ -319,7 +319,7 @@ feature 'safe-unicode', 'Sanitizes inputs with Unicode text.' => sub {
     # Note: Perl 5.8.1 bundles version 0.23.
     # Note: Perl 5.10.1 bundles 1.03 (per Unicode 5.1.0).
     requires 'Unicode::Normalize', '>= 1.03';
-    requires 'Unicode::UTF8', '>= 0.58';
+    requires 'Unicode::UTF8', '>= 0.60';
 };
 
 on 'test' => sub {
diff --git a/src/lib/Sympa/Tools/Text.pm b/src/lib/Sympa/Tools/Text.pm
index a8258e378..ef41927d7 100644
--- a/src/lib/Sympa/Tools/Text.pm
+++ b/src/lib/Sympa/Tools/Text.pm
@@ -307,8 +307,10 @@ sub guessed_to_utf8 {
 
     my $utf8;
     if ($Unicode::UTF8::VERSION) {
-        $utf8 =
-            eval { Unicode::UTF8::decode_utf8($text, Encode::FB_CROAK()) };
+        $utf8 = Unicode::UTF8::decode_utf8($text)
+            if Unicode::UTF8::valid_utf8($text);
+    } else {
+        $utf8 = eval { Encode::decode_utf8($text, Encode::FB_CROAK()) };
     }
     unless (defined $utf8) {
         foreach my $charset (map { $_ ? @$_ : () } @legacy_charsets{@langs}) {