As this article from OUT-LAW.com points out, previously some ISPs have steadfastly refused to block content flowing over their networks.
In part this has been to avoid legal responsibility for the delivery of undesirable or malicious code. However, AT&T's proposal seems not to be a security measure per se -- the stated reason is to monitor and block copyright infringement.
AT&T propose a "technology-based solution" -- however:
"Digital rights activists claim that filters do not take account of the exceptions from copyright legislation for fair use, including those for news reporting, criticism and review and educational use," the article says.
Is it possible to build filters clever enough to recognise "fair use" of content?