Ninth Circuit Screws Up 47 USC 230–Fair Housing Council v.

By Eric Goldman

Fair Housing Council of San Fernando Valley v., LLC, CV-03-09386-PA (9th Cir. May 15, 2007)

Just a couple months ago, in Perfect 10 v. CCBill, the Ninth Circuit issued an incredibly expansive 230 ruling. Today, in a highly fractured opinion, they go in the completely opposite direction, creating a significant exception to 230’s coverage that’s bound to spur plenty of new unmeritorious and ill-advised lawsuits from plaintiffs. Why the 180? Such is life in the Ninth Circuit.

The Opinion

This case involves the long-running Fair Housing Act lawsuit against, a website that allows users to advertise openings for roommates (a type of online classified ads). provides structured data fields with pull-down menus so that advertising roommates can provide consistent data and searchers can easily sort and compare data.

On the surface, it seems like it would be hard for to claim 230 protection for the resulting data, because the literal words are provided by However, users select the words that describe their situation, so the words are really provided by the user but structured to ensure a more functional database. Further, in the Carafano case, the Ninth Circuit had given a free pass to another website operator that allowed users to select among structured data fields, so it seemed like there was directly defense-favorable precedent on point.

In this ruling, the Ninth Circuit found otherwise, based on a confusing part of the statute. Per 230, ICSs aren’t liable for content provided by another ICP, and an ICP is defined as someone responsible in whole or in part for the content’s development. There are two ways to read this part of the statute:

* Reading #1: So long as the ICS is responsible in part (any part, even 0.1%) for the development of the content, then the ICS is an ICP and 230 isn’t available. This reading isn’t very useful because it would apply whenever an ICS edited any third party content, which is exactly what 230 routinely has been held to protect.

* Reading #2: So long as any third party ICP was responsible in part for the content’s development (even 0.1%), the ICS isn’t liable for it. This means that the ICS could have a great deal of involvement in the content but still avoid liability. This is by far the dominant interpretation of the statute.

Here, the Ninth Circuit interprets 230 using Reading #1. Thus, where provides pull-down menus, it loses 230 protection for the users’ responses. The court distinguishes the Carafano precedent by saying that the prankster in Carafano abused the system, while’s users used it as designed:

Carafano provided CDA immunity for information posted by a third party that was not, in any sense, created or developed by the website operator—indeed, that was provided despite the website’s rules and policies. Id. We are not convinced that Carafano would control in a situation where defamatory, private or otherwise tortious or unlawful information was provided by users in direct response to questions and prompts from the operator of the website.

Kozinski points to compiling search results as another way manages its data. allows users to search for certain types of data, such as women seeking women roommates. Kozinski says “By categorizing, channeling and limiting the distribution of users’ profiles, Roommate provides an additional layer of information that it is “responsible” at least “in part” for creating or developing.”

But Kozinski’s view of “partial responsibility for content development” is highly expansive. He says “we do not read [Carafano] as granting CDA immunity to those who actively encourage, solicit and profit from the tortious and unlawful communications of others.” Thus, he gives an example where a website actively encourages users to provide a target’s personal information as a way of getting even. Per Kozinski, this website wouldn’t get 230 coverage even if it requested certain categories of information but left the fields open (i.e., not using pull-down menus).

But, where provides an open field for “additional comments,” gets 230 protection. So open-ended Qs appear to be OK, but specific Qs aren’t. This is significant for the case on remand, as it appears that the most objectionable comments were in the “additional comments” field, and those will be excluded from consideration. However, the net result is that the case is going back on remand to determine if the pull-down menus and search functionality (now that 230 doesn’t apply to those) violate the substantive provisions of the Fair Housing Act.

Reinhardt issues a concurrence/dissent saying that should lose all 230 protection. He restates Kozinski’s opinion to hold that 230 can be lost either when “Roommate categorizes, channels and limits the distribution of information, thereby creating another layer of information,” or “Roommate actively prompts, encourages, or solicits the unlawful information.” He thinks the additional comments field satisfies both standards.

Ikuta concurs in part, saying that Kozinski’s active encouragement example was unnecessary, and points to the expansive 230 readings in Carafano and Batzel to point out that the precedent has defined ICP narrowly (i.e., more like reading #2).


Yuck–what a hairball. With 3 largely inconsistent opinions from a 3-judge panel, and a lead opinion that significantly narrows a fairly clear and relatively on-point Ninth Circuit precedent (Carafano), this case seems ripe for an en banc rehearing. I could even see this case going before the Supreme Court (especially because it conflicts with so many other precedents, including Ninth Circuit precedent), although I think the pull-down menus would be a lousy test case for the Supreme Court.

Meanwhile, if the Ninth Circuit doesn’t correct this opinion, and fast, I predict the following:

1) Websites will shy away from gathering structured data from users. This is silly, of course, because structured data can be more useful for users, but this ruling makes structured data much more risky.

2) Craigslist will lose its very similar case in the Seventh Circuit. The Seventh Circuit already had some bad 230 dicta in Doe v. GTE, and this opinion will give the Seventh Circuit judges all the ammo they need to hold Craigslist liable.

3) Plaintiffs are going to have a field day with the language in this opinion, especially as Reinhardt reformulated the test. I predict lots of new lawsuits probing this opinion.

In particular, this opinion can be read that any time a website reconstructs user data through its search engine, it loses 230 for that reconstruction. Read that way, this opinion could signal that Google and other search engines have no 230 protection for their search results. I trust this makes everyone nervous. If that issue ever made it back to the Ninth Circuit, I’m 100% convinced that Kozinski would clarify this opinion to provide 230 coverage to the search engines, However, in the interim, lots of plaintiffs and lower court judges could read this opinion differently, and that’s not going to be very fun for defense lawyers or their clients.

All in all, this case is a very unfortunate Cyberlaw development. Let’s hope the damage gets reversed pronto.

UPDATE: Commentary on the opinion:

* Eugene Volokh tries to make sense of this mess. The fact that he and I both can’t really figure out what the court said (and who agreed with what) isn’t a good sign.

* Laura Quilter makes a number of good points, including “I imagine that this case will be used by all who seek to limit Section 230’s broad immunity for ISPs. Whether the case constitutes a high-water mark for limitations on that liability, or merely a beachhead, remains to be seen.”

* Aaron Perzanowski, in a post entitled “Ninth Circuit Breaks Internet”: “The court, in three separate opinions, offered a novel reading of the statute that, unless limited by future panels to the unique facts of this case, will prove a font of disastrously uncertain liability for websites and other web-based services….Hopefully, future Ninth Circuit panels interpreting 230 are as quick to ignore precedent as this one was.”

* Joe Gratz is now the proud owner of “,” the domain used in Kozinski’s misguided hypothetical. He’s soliciting your suggestions of what to do with it.