Let’s now consider how these principles might apply to state regulations of social media platforms, and in particular to statutes (whether framed as public accommodations statutes or quasi-common-carrier statutes) that ban discrimination based on various attributes of a user or of the user’s speech. We’ll begin with two relatively simple and narrow hypotheticals, and then turn in a later post to the more controversial and ambitious statutes aimed at forbidding certain kinds of political discrimination, such as the ones recently enacted in Florida and Texas.
Antidiscrimination Statutes: Status
Federal public accommodations law likely doesn’t cover social media platforms, because it’s limited to only a few kinds of establishments. It also only bans discrimination based on race, religion, and national origin. But many states ban public accommodation discrimination in many establishments, based on many criteria.
Of course, today’s major social media platforms likely wouldn’t expressly exclude members based on, say, race or sex or sexual orientation. But consider a peculiar form of antidiscrimination law: bans on discrimination in places of public accommodation based on arrest or conviction history. Ann Arbor categorically bans such discrimination based on arrest record. Madison, Urbana, and Champaign do the same as to arrest record or conviction record. Connecticut bans discrimination based on expunged criminal records. New Jersey bans discrimination based on criminal history involving possession, distribution, or manufacturing of marijuana and hashish. Illinois, Hawaii, New York, and Wisconsin also ban such discrimination in employment, so it’s easy to imagine one of those states extending the ban to public accommodations.
Indeed, say Wisconsin is, rightly or wrongly, persuaded to do that (especially given Madison’s step in that direction). But say some social network—call it SafeBook—decides to ban people with a history of criminal offenses from portions of its site that children can visit. (Maybe sex offenses, but maybe also drug offenses; many parents might not want their children to fall in with the wrong crowd online and be exposed to bad influences.)
And say that Wisconsin courts conclude that the Wisconsin law applies to social media networks. This isn’t implausible: Some courts have already held that bans on disability discrimination in places of public accommodation apply to websites. The Wisconsin law also provides that the term “public place[s] of accommodation or amusement” “shall be interpreted broadly to include, but not be limited to, places of business or recreation,” “and any place where . . . amusement, goods, or services are available either free or for a consideration.” Finally, say that courts conclude that this nondiscrimination rule doesn’t violate the social media networks’ First Amendment rights, and isn’t preempted by § 230.
The Dormant Commerce Clause shouldn’t invalidate this law, at least if certain assumptions about geolocation (more on them shortly) are satisfied. This is especially clear if the law only protects the right of users in Wisconsin to interact with other users in Wisconsin (let’s call this Option 1). Just as a theme park in Wisconsin can’t exclude visitors with a criminal record, so SafeBook can’t keep a Wisconsinite from logging on and having online conversations with other Wisconsinites who have criminal records.
To be sure, Wisconsin’s law would have extraterritorial effects: If SafeBook is headquartered in, for instance, Kansas, presumably SafeBook will have to do many things in Kansas to comply. But of course any corporation that deals with customers all over the country through bricks-and-mortar stores would have to comply with the laws of those places where it operates. Likewise, any corporation that mails material to customers, or deals with them through phone calls, would have to comply with the laws of those places (for instance, in deciding what it must do to legally record phone calls with customers).
This analysis assumes what is increasingly commonplace: that SafeBook has access to geographical identification tools at a reasonable cost that can determine whether a user and the users with whom he is corresponding are in Wisconsin. So long as Wisconsin only requires SafeBook to use reasonable best efforts with geolocation tools—rather than, say, imposing strict liability for any criminal history discrimination against users who happen to be in Wisconsin, even if they appear to be coming from Iowa—then SafeBook can still maintain its criminal-offender-free experience for users in other states. It could just hide any Wisconsin criminal-offender users so users in other states can’t correspond with them, but show them to their fellow Wisconsinites.
This sort of geography-based variation in experience is similar to what the Ninth Circuit ruled that CNN could provide to accommodate disabled users in California. Safebook would have to go beyond CNN in at least one respect: It would have to take note of the place from which the item was posted, and not just, as in CNN’s case, the place from which the item was accessed. But that move might be technically easy, to the extent that Safebook can geolocate the poster when the post is put up, and then store that location information together with the other fields in the post, such as the post date, time, author, and text. Armed with that information, the SafeBook software can make sure that Wisconsinites will see posts from other Wisconsinites, even if it wants to continue to block posts by criminal offenders from or to other states.
To be sure, the challenge for SafeBook could become more complicated as more states impose such regulations. Some jurisdictions ban arrest and conviction record discrimination and some don’t. Some might ban such discrimination more or less broadly (again, recall New Jersey law, which currently bans only discrimination based on a history of having possessed, sold, or manufactured marijuana or hashish). Some ban sexual orientation discrimination and gender identity discrimination and some don’t. Some ban marital status discrimination and some don’t. Most ban religious discrimination but a few don’t. A few ban discrimination based on veteran status or military status but most don’t. SafeBook might need to survey the state laws and have different rules for different states, especially if it wants to institute other forms of discrimination, but even if it limits itself to discrimination based on criminal history.
But welcome to the American federal system, where companies that do business with people who are in multiple states must comply with the laws of those multiple states. Mail-order retailers, for instance, have to comply with the often byzantine tax rules of many states, even though “[s]tate taxes differ, not only in the rate imposed but also in the categories of goods that are taxed and, sometimes, the relevant date of purchase.” But that by itself doesn’t immunize the retailers from complying with state laws via readily available geolocation and other software tools.
Likewise for large social media platforms, at least to the extent that they can use geo-identification tools to achieve a reasonable level of compliance with multiple state laws as described above. These are firms, after all, whose businesses are all about writing software to cope with business opportunities and challenges, including software that for business-enhancement reasons treats users differently based on geography.
As we noted before, such compliance can be burdensome, and perhaps unduly burdensome, for small companies, whether small retailers or small platforms. The Dormant Commerce Clause balancing test (the Pike v. Bruce Church, Inc. test), under which a regulation may be struck down if “the burden imposed on such commerce is clearly excessive in relation to the putative local benefits,” might limit such state laws in some measures, as applied to small retailers. In the tax context, Wayfair recognized that the “burdens [of having to collect state sales tax] may pose legitimate concerns in some instances, particularly for small businesses that make a small volume of sales to customers in many States,” and that it may be significant that a state “affords small merchants a reasonable degree of protection,” for instance if they do very little business in the state.
Yet even when discussing small businesses facing multijurisdictional legal burdens, the Court noted that, “[e]ventually, software that is available at a reasonable cost may make it easier for small businesses to cope with these problems.” The Court appeared to be referring to multistate tax compliance software that is a cousin of the geolocation software we have highlighted. And any concern about undue burdens on interstate commerce should be further significantly reduced for large businesses, with users and advertisers all over the country, that deploy geo-identification software to serve business interests.
Now of course a harder question would arise if Wisconsin law insisted that SafeBook, as a condition of doing any business in Wisconsin, let any user from Wisconsin sign on (regardless of criminal history) and have conversations with all users on the platform (including, by implication, those from other states such as, say, Iowa). Let’s call this Option 2; it would go beyond regulating the experience of Wisconsin users of SafeBook and influence the experience of users in Iowa, who would end up interacting with some people with criminal histories (not fellow Iowans, who still wouldn’t be protected by Option 2, but Wisconsinites), simply because Wisconsin law so mandates.
Option 2 is harder because one can conceptualize the Wisconsin regulation as doing more than raising SafeBook’s costs of doing business in Iowa; the Wisconsin law can also be viewed as requiring Safebook to provide a certain form of online experience in Iowa that SafeBook wouldn’t otherwise provide. Option 2 for this reason moves in the direction of the Supreme Court’s price affirmation cases, which struck down state price affirmation laws basically because they mandated certain behavior in other states.
But, as noted above, the continuing validity of these cases is in question; and the argument that Option 2 is consistent with the Dormant Commerce Clause is powerful. Wisconsin has an interest in making sure that Wisconsinites are treated equally by places of public accommodation without regard to arrest or conviction records. That suggests that someone using Safebook in Madison should be entitled to have the same online experience—including the same conversations with out-of-staters—regardless of whether he has, say, a marijuana conviction on his record. Wisconsin would be regulating the experience that SafeBook is providing for people who are visiting SafeBook from Wisconsin, even though in the process it would incidentally also affect the experience of SafeBook visitors from outside Wisconsin.
An analogy might be a physical delivery service in Wisconsin that refused to accept packages—including for interstate shipment—from people who had sex crime convictions (perhaps because the owner just didn’t want to do business with people who had committed such heinous acts). The Wisconsin Legislature might well conclude that this is improper discrimination against Wisconsin residents and enact an Option-2-like public accommodation statute to forbid such discrimination, even though such a statute would also affect the delivery service’s actions in delivering packages from Wisconsin to other states.
Let us then add one more twist: Say that, while Wisconsin bans discrimination based on criminal history, North Carolina requires social media platforms to exclude people with certain kinds of criminal history from portions of social media platforms that are targeted to children. This would indeed put SafeBook in a difficult position: If it keeps a Wisconsin user with a particular history from interacting with North Carolina users who are accessing a particular portion of the platform, then it would be violating Wisconsin law (at least in Option 2). If it allows the Wisconsin user to interact with the North Carolina users, then it would be violating North Carolina law.
This sort of actual inconsistency might justify keeping one or the other law from applying in those situations under Pike balancing. We doubt, though, that the mere hypothetical possibility of such inconsistency should categorically foreclose antidiscrimination laws from applying to social media platforms.
. 42 U.S.C. § 2000a; Lewis v. Google LLC, No. 20-16073, 2021 WL 1423118 (9th Cir. Apr. 15, 2021).
. 42 U.S.C § 2000a.
. Ann Arbor (Mich.) Code of Ordinances § 9:151.
. Madison (Wisc.) Code of Ordinances § 39.03(5); Urbana (Ill.) Code of Ordinances § 12-39, -63; Champaign (Ill.) Code of Ordinances § 17-3, -56.
. 2021 Conn. Legis. Serv. P.A. 21-32, sec. 23 (S.B. 1019) (starting Jan. 1, 2023).
. N.J. Stat. Ann. § 10:5-50.
. Ill. Comp. Stat. Ann. §§ 5/2-103 to -103.1; Haw. Rev. Stat. Ann. § 378-2; N.Y. Exec. L. §§ 296.15–.16; Wisc. Stats. Ann. § 111.335.
. See Winegard v. Crain Commc’ns, Inc., No. 20-CV-01509 (AJN), 2021 WL 1198960 (S.D.N.Y. Mar. 30, 2021); Sullivan v. BDG Media, Inc., 71 Misc. 3d 863 (N.Y. Sup. Ct. 2021); Nat’l Fed’n of the Blind v. Scribd Inc., 97 F. Supp. 3d 565 (D. Vt. 2015); Nat’l Ass’n of the Deaf v. Netflix, Inc., 869 F. Supp. 2d 196 (D. Mass. 2012); see also Sullivan v. Study.com LLC, No. 18-CV-1939 (JPO), 2019 WL 1299966 (S.D.N.Y. Mar. 21, 2019) (likewise, but noting that the defendant did not contest being treated as a public accommodation). For more on whether web sites that are ancillary to the sale of other goods and services are places of public accommodations for purposes of federal disability discrimination law, see Gil v. Winn-Dixie Stores, Inc., 993 F.3d 1266, 1277 (11th Cir. 2021) (saying no); Robles v. Domino’s Pizza, LLC, 913 F.3d 898, 905–06 (9th Cir. 2019) (saying yes); Blake E. Reid, Internet Architecture and Disability, 95 Ind. L.J. 591, 597–99 (2020) (summarizing the dispute); Johanna Smith & John Inazu, Virtual Access: A New Framework for Disability and Human Flourishing in an Online World, 2021 Wisc. L. Rev. 719, 759–65 (2021).
. Wisc. Stat. § 106.52(1)(e)1. Though the law excludes private clubs, that exclusion is limited to “bona fide private, nonprofit organization[s]” that aren’t open to the public generally. Wisc. Stat. § 106.52(1)(e)2.
. See Eugene Volokh, Treating Social Media Platforms Like Common Carriers?, 1 J. Free Speech L. 377 (2021).
. See Adam Candeub & Eugene Volokh, Interpreting 47 U.S.C. § 230(c)(2), 1 J. Free Speech L. 175 (2021). Note also that § 230(c)(2) preempts liability for “any action voluntarily taken in good faith to restrict access to or availability of material that the provider or user considers to be obscene, lewd, lascivious, filthy, excessively violent, harassing, or otherwise objectionable, whether or not such material is constitutionally protected”—it doesn’t discuss action to block access based on a person’s identity, rather than based on the content of material.
. We use “Wisconsinite” here as shorthand for people who are in Wisconsin, or more precisely people who appear to be in Wisconsin based on geolocation tools. We’re not focusing here on place of residence or citizenship, which would be impractical to determine—just as, for instance, the Ninth Circuit in the CNN case was focusing on the user’s location and not the user’s residence or citizenship.
. Kearney v. Salomon Smith Barney, Inc., 137 P.3d 914 (Cal. 2006)
. To be sure, that means that SafeBook couldn’t just delete posts by people with criminal histories, but would have to keep them on file and show them to Wisconsinites while concealing them from others. But that would be a very easy algorithm to write; and any extra disk space that would be required to keep the material that would otherwise be deleted would be quite inexpensive given today’s technology.
. Alternatively, if state law dictates that the authors’ location be determined based on where the authors are located when they subscribe to the service—rather than based on where they are when they post each item—then Safebook could just store the location in the author’s user record, rather than once for each post.
. If it’s technically too difficult, then a court could presumably conclude that the Wisconsin requirement is unconstitutional under the Pike balancing test, see infra text accompanying note 146; but there’s no reason to assume a priori that this would be too difficult.
. See generally Nat’l Conf. of State Legs., State Public Accommodation Laws, https://www.ncsl.org/research/civil-and-criminal-justice/state-public-accommodation-laws.aspx [https://perma.cc/8L6G-5MCS].
. South Dakota v. Wayfair, Inc., 138 S. Ct. 2080, 2098 (2018).
. 397 U.S. 137, 142 (1970); see Greater L.A. Agency on Deafness, 742 F.3d at 433 (applying the Pike test); Goldsmith & Sykes, supra note 11, at 806 (discussing a possible Pike-based analysis in extraterritoriality cases).
. See Kearney v. Salomon Smith Barney, Inc., 45 Cal. Rptr. 730, 739 (Cal. 2006) (“On its face, application of the California law here at issue would affect only a business’s undisclosed recording of telephone conversations with clients or consumers in California and would not compel any action or conduct of the business with regard to conversations with non-California clients or consumers. Although SSB may attempt to demonstrate, at a later stage in the litigation, that application of the California statute would pose an undue and excessive burden on interstate commerce by establishing that it would be impossible or infeasible for SSB to comply with the California statute without altering its conduct with regard to its non-California clients and that the burden that would be imposed upon it ‘is clearly excessive in relation to the putative local benefits’ (Pike v. Bruce Church, Inc.), SSB clearly cannot prevail on such a theory at the demurrer stage of the proceeding.”) (citation trimmed).
. 138 S. Ct. at 2098.
. See Brief of South Dakota, https://www.supremecourt.gov/DocketPDF/17/17-494/36735/20180226222258706_17-494%20ts.pdf [https://perma.cc/QKJ6-Y74V], at 14-15; Oral Argument Transcript, https://www.oyez.org/cases/2017/17-494 [https://perma.cc/B3M8-ZLFK].
 See, e.g., Healy v. Beer Institute, Inc., 491 U.S. 324, 336–38 (1989).
. Assume the delivery service isn’t governed by federal common carrier laws, which would presumably independently ban the service from imposing such conditions.
. Assume also that the North Carolina law is upheld against a First Amendment challenge because it’s narrower than the overbroad ban struck down in Packingham v. North Carolina, 137 S. Ct. 1730 (2017). See id. at 1743 (Alito, J., concurring in the judgment) (suggesting that narrower restrictions on “an adult previously convicted of molesting children from visiting a dating site for teenagers” or “a site where minors communicate with each other about personal problems” might be constitutional).
Add a Comment