$_ = do {my $t  = "";
            $t .= defined $1 ? do {(my $t = $1) =~ y/\\//d; $t} : $2
                  while /((?:[^\\]+|\\(?!\*))+)|(\\\*(?:[^*]+|\*(?!\\)
+)+\*\\)/g;
            $t
};
[download]

Japhy can probably tell you how to do it in one go, with much looking about, but I'll have to settle for three lines: first remove and store comments, then strip the backslashes, then restore the comments. This seems to work:

my $test = q|\AFZ\BJK \*\AFZ...*\ |;
my @comments;
$test =~ s/(\\\*.*?\*\\)/ push @comments, $1; '%%comment%%' /eg;
$test =~ s/\\//g;
$test =~ s/%%comment%%/ shift @comments /eg;
print $test;
[download]

Any old marker will do as long as it doesn't appear in your text: something long and random would probably be best. %%comment%% was just the first thing that came into my head.

ps. goes clunk, i know, but easy to follow

how would i make this work if the comment spanned more than 1 line?

just adding an s modifier to each regex ought to do it: it'll then treat the whole thing as if it were one line (or, strictly, it'll treat line-endings as just another character), as the early parts of perlre will explain. I should have put that in, but your example was just one line so I didn't bother:

$test =~ s/(\\\*.*?\*\\)/ push @comments, $1; '%%comment%%' /egs;
$test =~ s/\\//gs;
$test =~ s/%%comment%%/ shift @comments /egs;
[download]

$test=q|\AFZ\BJK \*\AFZ...*\ |;
$new.=$1 while $test=~/([^\\]|\\\*.*?\*\\)/g;
[download]

$test=q|\AFZ\BJK \*\A\*F*\Z...*\ |;
$comment=qr/\\\*(?:[^\\]*|\\[^*]|(??{$comment}))*?\*\\/;
$new.=$1 while $test=~/([^\\]|$comment)/g;
[download]

$test ="\\AFZ\\BJK \\\*\\AFZ...\*\\";
$span = 0; # Span is a span counter
foreach( split(/(\\\*|\*\\)/,$test) ){ # Split on spans keeping the co
+mmenttags
    if($_ eq '\\*'){ # See if its a comment-start
        $span ++;
    } elsif($_ eq '*\\'){ # if it's a comment-end
        $span --;
    } elsif($span eq 0){ # Only if not in comment
        $_=~s/\\//igs;
    }
    $newtest.=$_;
}
[download]