If you encode UTF-8 strings into perl's internal format, you don't need that kind of hack:

use Encode qw(encode decode);
my $str = # some input method here;
$str = encode("utf8", $str);
# do your pattern matching here

print decode("utf8", $str);
[download]

Or look

If you have non-ascii characters in the source of your script, use utf8;. And take a look at perluniintro and perlunicode.

Or in what way do utf-8 strings "break" without the string eval?

And we may give you a much better answer if you told us what you're trying to achieve with that regexp — few here understand what \x(w{2}) is supposed to mean or why that w had to be escaped inside the eval.

Avoid the XY problem.