A valid cut is one that isn't followed by a continuation byte (0b10xx_xxxx).

sub truncate_utf8 {
   my ($utf8, $len) = @_;
   $len += 0;  # Make sure $len is a number.
   return $utf8 =~ s/^.{0,$len}\K(?![\x80-\xBF]).*//sr;
}
[download]

sub truncate_utf8 {
   my ($utf8, $len) = @_;
   return $utf8 if length($utf8) <= $len;

   my $next = substr($utf8, $len, length($utf8)-$len, '');
   $next = chop($utf8) while (ord($next) & 0xC0) == 0x80;
   return $utf8;
}
[download]

Both of these take text that is already encoded using UTF-8.

Update: Fixed typo mentioned by haukex.
Update: Made clear what the input should be.