Appeals Court Issues Strong CDA 230 Ruling, But It Will Be Misleadingly Quoted By Those Misrepresenting CDA 230
from the mostly-good,-but-a-bit-of-bad dept
Last Friday, the DC circuit appeals court issued a mostly good and mostly straightforward ruling applying Section 230 of the Communications Decency Act (CDA 230) in a perfectly expected way. However, the case is notable on a few grounds, partly because it clarifies a few key aspects of CDA 230 (which is good), and partly because of some sloppy language that is almost certainly going to be misquoted and misrepresented by those who (incorrectly) keep insisting that CDA 230 requires “neutrality” by the platform in order to retain the protections of the law.
Let’s just start by highlighting that there is no “neutrality” rule in CDA 230 — and (importantly) the opposite is actually true. Not only does the law not require neutrality, it explicitly states that it’s goal is for there to be more content moderation. The law explicitly notes that it is designed:
to remove disincentives for the development and utilization of blocking and filtering technologies that empower parents to restrict their children?s access to objectionable or inappropriate online material
In short, the law was designed to encourage moderation, which, by definition, cannot be “neutral.”
Now, onto the case. It involved a bunch of locksmiths who claim that there are “scam locksmiths” claiming to be local in areas they are not, and the various search engines (Google, Microsoft and Yahoo are the defendants here) are putting those fake locksmiths in their search results, meaning that the real locksmiths have to spend extra on advertising to get noticed above the scam locksmiths.
You might think that if these legit locksmiths have been wronged by anyone, it’s the scam locksmiths, but because everyone wants to blame platforms, they’ve sued the platforms — claiming antitrust violations, false advertising, and a conspiracy in restraint of trade. The lower court, and now the appeals court, easily finds that Section 230 protects the search engines from these claims, as the content from the scam locksmiths is from the scam locksmiths, and not from the platforms.
The attempt to get around CDA 230 is mainly by focusing on one aspect of how the local search services work — in that most of the services try to take the address of the local business and place a “pin” on a map to show where the business is physically located. The locksmiths argue that creating this map and pin involves content that the search engines create, and therefore it is not immune under CDA 230. The appeals court says… that’s not how it works, since the information is clearly “derived” directly from the third parties:
The first question we must address is whether the defendants? translation of information that comes from the scam locksmiths? webpages — in particular, exact street addresses — into map pinpoints takes the defendants beyond the scope of ? 230 immunity. In considering this question, it is helpful to begin with the simple scenario in which a search engine receives GPS data from a user?s device and converts that information into a map pinpoint showing the user?s geographic location. The decision to present this third-party data in a particular format — a map — does not constitute the ?creation? or ?development? of information for purposes of ? 230(f)(3). The underlying information is entirely provided by the third party, and the choice of presentation does not itself convert the search engine into an information content provider. Indeed, were the display of this kind of information not immunized, nothing would be: every representation by a search engine of another party?s information requires the translation of a digital transmission into textual or pictorial form. Although the plaintiffs resisted this conclusion in their briefs, see Locksmiths? Reply Br. 3 (declaring that the ?location of the inquiring consumer . . . is determined entirely by the search engines?), they acknowledged at oral argument that a search engine has immunity if all it does is translate a user?s geolocation into map form, see Recording of Oral Arg. at 12:07-12:10.
With this concession, it is difficult to draw any principled distinction between that translation and the translation of exact street addresses from scam-locksmith websites into map pinpoints. At oral argument, the plaintiffs could offer no distinction, and we see none. In both instances, data is collected from a third party and re-presented in a different format. At best, the plaintiffs suggested that a line could be drawn between the placement of ?good? and ?bad? locksmith information onto the defendants? maps. See id. at 12:43-12:58 (accepting that, ?to the extent that the search engine simply depicts the exact information they obtained from the good locksmith and the consumer on a map, that appears to be covered by the [Act]?). But that line is untenable because, as discussed above, Congress has immunized the re-publication of even false information.
That’s a nice, clean ruling on what should be an obvious point. But having such clean language could be useful for citations in future cases. It is notable, at least, (and useful) that the court clearly states: “Congress has immunized the re-publication of even false information.” Other courts have made this clear, but having it in such a nice, compact form that is highly quotable is certainly handy.
There are a few other attempts to get around CDA 230 that all fail — including using the fact that the “false advertising” claim is under the Lanham Act (which is often associated with trademark law), and CDA 230 explicitly excludes “intellectual property” law. But that doesn’t magically make the false advertising claims “intellectual property,” nor does it exclude them from CDA 230 protections.
But, as noted up top, there is something in the ruling that could be problematic going forward concerning the still very incorrect argument that CDA 230 requires the platforms be “neutral.” The locksmiths’ lawyers argued that even if the above case (of putting a pin on a map) didn’t make the search engines “content creators,” perhaps they were content creators when they effectively made up the location. In short: when these (and some other) local search engines don’t know the actual exact location of a business, they might put in what is effectively a guesstimate, usually placing it in a central location of an expected range. As the court explains:
The plaintiffs describe a situation in which the defendants create a map pinpoint based on a scam locksmith?s website that says the locksmith ?provides service in the Washington, D.C. metropolitan area? and ?lists a phone number with a ?202? area code.? Locksmiths? Br. 8; see also Locksmiths? Reply Br. 4-5. According to the plaintiffs, the defendants? search engines use this information to ?arbitrarily? assign a map location within the geographic scope indicated by the third party.
Legally, that does represent a slightly different question — and (if you squint) you can kinda see how someone could maybe, possibly, make the argument that if the local search engines take that generalized info and create a pin that appears specific to end users, that it has somehow “created” that content. But the court (correctly, in my opinion) says “nope,” and that since that pin is still derived from the information provided by a 3rd party, 230 protects. This is good and right.
The problem is that the court went a bit overboard in using the word “neutral” in describing this, using the word in a very different way than most people mean when they say “neutral” (and in a different way than previous court rulings — including those cited in the case — have used the word neutral):
We conclude that these translations are also protected. First, as the plaintiffs do not dispute, the location of the map pinpoint is derived from scam-locksmith information: its location is constrained by the underlying third-party information. In this sense, the defendants are publishing ?information provided by another information content provider.? Cf. Kimzey v. Yelp!, Inc., 836 F.3d 1263, 1270 (9th Cir. 2016) (holding that Yelp?s star rating system, which is based on receiving customer service ratings from third parties and ?reduc[ing] this information into a single, aggregate metric? of one to five stars could not be ?anything other than usergenerated data?). It is true that the location algorithm is not completely constrained, but that is merely a consequence of a website design that portrays all search results pictorially, with the maximum precision possible from third-party content of varying precision. Cf. Carafano v. Metrosplash.com, Inc., 339 F.3d 1119, 1125 (9th Cir. 2003) (?Without standardized, easily encoded answers, [Matchmaker.com] might not be able to offer these services and certainly not to the same degree.?).
Second, and also key, the defendants? translation of thirdparty information into map pinpoints does not convert them into ?information content providers? because defendants use a neutral algorithm to make that translation. We have previously held that ?a website does not create or develop content when it merely provides a neutral means by which third parties can post information of their own independent choosing online.? Klayman, 753 F.3d at 1358; accord Bennett, 882 F.3d at 1167; see Kimzey, 836 F.3d at 1270 (holding that Yelp?s ?star-rating system is best characterized as the kind of neutral tool operating on voluntary inputs that . . . [does] not amount to content development or creation? (internal quotation marks omitted) (citing Klayman, 753 F.3d at 1358)). And the Sixth Circuit has held that the ?automated editorial act[s]? of search engines are generally immunized under the Act. O?Kroley v. Fastcase, Inc., 831 F.3d 352, 355 (6th Cir. 2016).
Here, the defendants use automated algorithms to convert third-party indicia of location into pictorial form. See supra note 4. Those algorithms are ?neutral means? that do not distinguish between legitimate and scam locksmiths in the translation process. The plaintiffs? amended complaint effectively acknowledges that the defendants? algorithms operate in this fashion: it alleges that the words and numbers the scam locksmiths use to give the appearance of locality have ?tricked Google? into placing the pinpoints in the geographic regions that the scam locksmiths desire. Am. Compl. ? 61B. To recognize that Google has been ?tricked? is to acknowledge that its algorithm neutrally translates both legitimate and scam information in the same manner. Because the defendants employ a ?neutral means? and an ?automated editorial act? to convert third-party location and area-code information into map pinpoints, those pinpoints come within the protection of ? 230.6
See all those “neutral means” lines? What the court means is really automated and not designed to check for truth or falsity of the information. It does not mean “unbiased,” because any algorithm that is making decisions is inherently and absolutely “biased” towards trying to choose what it feels is the “best” solution — in this case, it is “biased” towards approximating where to put the pin.
The court is not, in any way, saying that a platform need be “neutral” in how it applies moderation choices, but I would bet a fair bit of money that many of the trolls (and potentially grandstanding politicians) will use this part of the court ruling to pretend 230 does require “neutrality.” The only thing I’m not sure about is how quickly this line will be cited in a bogus lawsuit, but I wouldn’t expect it to take very long.
For what it’s worth, after I finished writing this, I saw that professor Eric Goldman had also written up his analysis of the case, which is pretty similar, and includes a few other key points as well, but also expects the “neutral means” line to be abused:
I can see Sen. Cruz seizing on an opinion like this to say that neutrality indeed is a prerequisite of Section 230. That would be less of a lie than his current claim that Section 230 only protects ?neutral public forums,? but only slightly. The ?neutrality? required in this case relates only to the balance between legal and illegal content. Still, even when defined narrowly and precisely like in the Daniel v. Armslist case, the term ?neutrality? is far more likely to mislead than help judges. In contrast, the opinion cites the O?Kroley case for the proposition that Section 230 protects ?automated editorial act[s],? and that phrasing (though still odd) is much better than the term ?neutrality.?
The overall ruling is good and clear — with just this one regrettable bit of language.