I think this is a crucial distinction, because efforts to filter end users' connections (as opposed to making them pay consequences for their actions after the fact) have always been controversial, even when the content is illegal. The Center for Democracy and Technology successfully overturned a Pennsylvania law that required ISPs to block overseas child pornography sites, partly on the grounds that the filtering included many third-party Web sites as collateral damage. I've argued that a similar private-sector initiative called Canada Cleanfeed, where Canadian ISPs attempt to block child pornography Web sites, would do more harm than good. On the other hand, nobody's fighting very hard for the cause of child pornography downloaders who were caught and arrested. Web sites get sued and shut down all the time, but it was bigger news when Canadian ISP Telus blocked the Web site of a Telus labor union for three days. So it's a big deal whether we're talking about "pre-emptive" filtering, or fighting piracy "reactively" by going after violators.
AT&T Senior VP James Cicconi said in e-mail that "discussion about what the technology will or won't do is premature until we can invent it", but most of the hints so far have been that the anti-piracy technology will be "pre-emptive", i.e. filtering users' connections. Cicconi said on a conference panel that AT&T has to spend billions on network maintenance to carry illegal pirated traffic -- which they probably couldn't recoup by suing people, so the only way to prevent that would be to block it. And Cicconi has referred to the technology several times as a "network-based solution" -- but what else could that mean, except filtering?
So let's assume that's what's on the horizon. Interestingly, Cicconi said that AT&T did not plan to block actual Web sites. However, he said in e-mail, "If one could, with a high degree of certainty, spot and isolate illegal traffic from an offshore site, would you not think the copyright holders would have a reasonable argument for a court order to block that traffic (as opposed to the site itself)?" Presumably this could refer to a Web page with an index of links to BitTorrent files -- so they'd be willing to block the BitTorrent links, but not the Web page? But from that point of view, why not just block Web sites too? If an overseas webpage has a list of links to pirated content, and that content is served over http from the same Web server, wouldn't they want to block it?
But I doubt this would stem much piracy in the long run, because connection filtering to fight piracy became more commonplace, then the next generation of p2p file-trading programs would all just have circumvention capabilities built into them, that let you route your connection through a friend at an unfiltered ISP. You're on AT&T, you upload a file to your friend on Verizon which earns you some "credits" with his node in the p2p network, and instead of redeeming those credits to download a file from him, you use his node as a proxy to download a file indirectly from a site in Russia that AT&T is blocking you from accessing. Advanced users can do this already with tools like Virtual Private Networks and Tor, and some tweaks in a p2p program would just bring it within the range of the casual user.
On the other hand, if AT&T starts filtering traffic, it could set a bad precedent that any time a party in a legal proceeding wants a site declared "illegal", they can demand that AT&T (or other ISPs) block the site. It could be a site libeling a person, or a site hosting a decryption tool that breaks some company's poorly-designed code, or pretty much anything that some powerful person wanted to go away. Meanwhile, if an AT&T customer did get accused of downloading pirated content, now they could invoke the "AT&T didn't stop me" defense -- they thought that AT&T was filtering illegal content, and if they could get to it, then that meant it was legal! In both cases the problem comes from someone using the argument that once AT&T started doing any filtering at all, they should have gone further.
So I would watch the situation closely, even if you're not an AT&T user, and don't assume the situation will take care of itself. Cicconi said, "If a company like ours does dumb things and upsets our customers, we will lose them to someone else," which is something I'm skeptical of whenever I hear it used to defend various draconian anti-spam measures, but in this case I think it's even less applicable. When you're talking about spam filters, at least they always bring some benefit to the user (less spam), and the question is whether the free market weighs those benefits properly against the costs (more lost mail). On the other hand, if an ISP filters the user's connection, that brings no benefit to the user, and in a truly efficient market, all customers of such an ISP would just switch to an unfiltered one -- if that doesn't happen, it simply means the market in that case is not efficient. Is your ISP filtering your connection right now? Probably not, but how could you tell if they were? Right now we assume that ISPs don't filter connections because generally it's "just not done" (except when it is). In a few years we might not be so sure.