Comments on Sen. Hawley’s “[Ending] Support for Internet Censorship Act”

Sen. Josh Hawley’s “[Ending] Support for Internet Censorship Act” (S. 1914) would require large(ish) Internet companies to prove to the Federal Trade Commission (FTC) that they publish user content in a politically neutral way–or categorically lose Section 230 immunity. In other words, a government censorship board could legally penalize a major Internet service for exhibiting any sort of political bias.

[Note: I’ve bracketed the word “Ending” because the title “Support for Internet Censorship Act” is more accurate.]

The bill has no redeeming features; no interesting policy ideas to consider for future proposals; no meritorious conversation-starters. Some of the bill’s worst aspects:

Political Bias Isn’t Measurable. The bill defines political bias:

The moderation practices of a provider of an interactive computer service are politically biased if—

‘‘(I) the provider moderates information provided by other information content providers in a manner that—

‘‘(aa) is designed to negatively affect a political party, political candidate, or political viewpoint; or

‘‘(bb) disproportionately restricts or promotes access to, or the availability of, information from a political party, political candidate, or political viewpoint; or

‘‘(II) an officer or employee of the provider makes a decision about moderating information provided by other information content providers that is motivated by an intent to negatively affect a political party, political candidate, or political viewpoint.

There are several major problems with this approach:

First, the bill promotes false equivalencies. Not all political parties and viewpoints are equally legitimate. For example, I don’t think the Republican Party and the American Nazi Party are equally legitimate (please, no “jokes” about how their views are converging); and I think most people would expect Internet services to treat American Nazi Party candidates and party platforms less favorably than the Republican Party’s candidates/platforms. Yet, the bill would prevent Internet services from doing that.

Second, the term “political viewpoint” is broad. Pretty much every topic people care about has partisan dimensions–e.g., Russia, voting, evolution, even vaccines. If being anti-vax is a political viewpoint, the bill says Internet services must treat pro-vaccine and anti-vaxxer content equally.

Even limiting the bill to political parties and their platforms wouldn’t solve the problem. Imagine the Pastafarians form a political party. They’d need to be treated the same as the Republican Party. Or as I mentioned previously (only partially in jest), perhaps scatologicalists will claim they have a political viewpoint and sue Internet services for discriminating against pissing videos.

Because it’s impossible to distinguish between “political viewpoints” and other viewpoints, the bill’s anti-discrimination provisions would make it impossible for any Internet service to moderate content at all (including both leave-up and removal decisions) without risking allegations of non-neutrality. In light of how much bad content is online despite the Internet services’ current best efforts, that prospect should terrify everyone.

Third, the “disproportionate” standard assumes a baseline for comparison. What is that baseline? Proportions could be measured across Internet services, e.g., Service X removes 40% of anti-vaxxer content and Service Y removes 45% of anti-vaxxer content. But which is the right baseline? If they disagree, both of them are allegedly biased, one more tolerant of anti-vaxxers and the other less tolerant, meaning neither is neutral. Or perhaps the formula would measure the number of removed/promoted pro- and anti- content on a topic divided by the total content on that topic. If the percentages aren’t identical, then there is bias. Apply that formula to thousands of political “viewpoints,” and the online service must have identical pro- vs. anti- percentages for each and every topic. How likely is that?

Fourth, a service also would be biased if any Internet service’s employee intends to negatively affect a party/candidate/viewpoint. Facebook has 30,000 people (a mix of employees and contractors, if that matters) working on Safety & Security. If any one of them acts with an intent to discriminate politically, then the entire company would categorically lose Section 230 immunity–unless the company makes a public disclosure and fires/disciplines the employee after learning of the discriminatory act (I’m not sure how this obligation interacts with employment law). There might be smoking gun evidence of impermissible intent. More likely, discriminatory intent is shown statistically, i.e., crunch each employee’s decisions on political topics and find the outliers. With a large enough workforce, there will be statistical “proof” that at least one employee intended to favor or punish some party/candidate/viewpoint.

Neutral Publishing Is an Oxymoron. UGC services are content publishers. The publication process inherently involves prioritizing some content over others. That’s not bias or discrimination, it’s “editorial discretion”–and it’s what we want and expect from every publisher. The bill contemplates that Internet services will publish user content “neutrally,” but the very act of publishing content is never neutral. By punishing publishers for acting like publishers, the bill ensures that every Internet service will fail the legal standard.

Taking Away Section 230 Is a Nuclear Option. The bill would categorically remove Section 230, across the full range of circumstances it applies, for violating a single pre-condition (the lack of political neutrality). Section 230 squelches a wide range of unmeritorious lawsuits. Without its protection, the litigation floodgates against Internet services would open up in a truly unprecedented manner. The law would be a nuclear remedy for a single type of problem.

The FTC Isn’t Well-Positioned to Be a Censorship Board. I imagine the FTC is unenthusiastic about this bill. Nothing about the bill plays to its strengths. The FTC is strongest at policing deviant marketplace behavior. They are lousy at policing editorial content (e.g., they struggle with distinguishing editorial content from covert advertising). The non-partisan FTC has no experience evaluating the legitimacy of political content. Further, users could “submit complaints or evidence [to the FTC] that they have been subject to politically biased content moderation by a provider of an interactive computer service.” Can you imagine the complaint volume, and associated workload, when every user aggrieved about a content moderation decision complains to the FTC? The bill’s requirements would swallow up massive amounts of the FTC’s capacity and mindshare, reducing their ability to do the essential work they’re best at.

The FTC Approval Standards Are Impossible to Satisfy. The FTC would let an Internet service retain Section 230’s immunity only if the Internet service proves “by clear and convincing evidence that the provider does not (and, during the 2-year period preceding the date on which the provider submits the application for certification, did not) moderate information provided by other information content providers in a politically biased manner.” The Internet service must prove the negative–that they didn’t discriminate–by “clear and convincing evidence,” and convince 4 of 5 commissioners (which necessitates at least 1 vote from appointees from each political party) that the Internet service satisfied this high evidentiary standard. Due to the “political neutrality” oxymoron and the statistical impossibility of showing that every political viewpoint was treated identically, no Internet service will successfully obtain the FTC’s approval.

The Bills Reaches Too Many, and the Wrong Kinds of, Services. The law would apply to “providers of interactive computer services” with 30M US monthly average users (MAUs), 300M worldwide MAUs, or $500M/yr global revenue. The law targets Google/YouTube, Facebook/Instagram/WhatsApp, and Twitter, but dozens of other services meet these thresholds. Apple, Microsoft (including LinkedIn and GitHub), Oath (including AOL, Yahoo, and Tumblr), Amazon (including Twitch), and eBay all surpass the revenue threshold, as would any major company with a UGC sideline (e.g., the New York Times with its reader comments). Some other services that meet the thresholds: Pinterest ($1B projected in 2019), Snap ($1.2B revenue in 2018), TripAdvisor ($1.6B in 2018 revenue), Yelp ($1B in 2018 revenue), Etsy ($600M in 2018 revenue), Dropbox ($1.4B in 2018 revenue), Box.com ($500M in 2018 revenue), WeChat (1.1B global MAU), TikTok (500M global MAU), Weibo (460M global MAU), Reddit (330M global MAU), Baidu (300M global MAU), and possibly Vimeo, Craigslist, Imgur, and Discord (ha ha). (Statista is my source of global MAU stats). This is clearly not a complete list. Please email me suggestions of other companies that clear the thresholds. Note: Wikimedia is a 501(c)(3), which the bill excludes.

Remember, ALL of these companies would automatically lose Section 230 protection unless they obtain the FTC’s license, which I just said none of them can do.

The law doesn’t distinguish between content hosts and other types of services. How would the FTC certify political neutrality for e-commerce services like Amazon, eBay, or Etsy? If the marketplaces remove the sale of Nazi paraphernalia, does that count against them for neutrality purposes? Similarly, what does political neutrality look like for consumer review sites? Yelp doesn’t allow political comments by non-customers (remember the Red Hen restaurant imbroglio?), but customers can submit political comments that might be irrelevant or defamatory. What should Yelp do? And how could a traditional media outlet like the New York Times or Fox News that enables UGC qualify as politically neutral? Political neutrality is nonsensical, but it’s especially nonsensical when applied outside the law’s paradigmatic target of mass-market global social media services.

Constitutional Problems. The bill conditions an exceptionally valuable legal immunity on political neutrality. Stated differently, any Internet service that has any political slant (intentionally or not) will be treated legally worse than a politically neutral service. Thus, for example, a service that started out with a political slant and rode it to success would need to immediately change its editorial standards upon reaching the law’s quantitative thresholds or the FTC will deny its Section 230 license. Framed this way, the bill engages in viewpoint-based discrimination, i.e., if an Internet service has any political viewpoint, then the law discriminates against it compared to those without a political viewpoint. This ought to trigger strict scrutiny, and I don’t see how the bill could survive that. In a similar situation, the Supreme Court in Iancu v. Brunetti struck down a law that conditioned a government privilege (trademark registration) on the speaker’s viewpoint.

In effect, the bill creates a licensing scheme for Internet publishers–a broadcasting-style regulation of the Internet. In 1997, the Supreme Court (in Reno v. ACLU) said the Internet wasn’t like broadcasting. As Justice Stevens eloquently said, “our cases provide no basis for qualifying the level of First Amendment scrutiny that should be applied to this medium.” That precedent reinforce the bill’s unconstitutionality.

Conclusions

This bill reminded me of my reaction to FOSTA’s introduction. Both times, I felt the drafters didn’t understand Section 230’s “hydraulics” and sought to destroy Section 230 for some unrelated political objective.

Unlike FOSTA, the bill has been universally panned publicly. Only a few outliers have publicly supported this bill (a move that permanently costs them any credibility with me). I don’t see how this bill can advance.

Here’s my puzzle: why didn’t Sen. Hawley propose to categorically end Section 230 when companies reach the numerical thresholds? [To state the obvious, I would vigorously oppose such a proposal.] After all, if companies have little chance actually getting the FTC license to keep Section 230, the net result would be the same; and a categorical exclusion would save everyone time and money and avoid the worst constitutional angles. As far as I can tell, “political neutrality” was essential to the bill’s messaging. However, seeing the negative reaction from communities I’d have expected to sympathize, I can’t tell if Sen. Hawley misjudged the audience, decided to appeal only to a tiny niche, or is playing some super-strategic 4-dimensional chess that I don’t understand.

Sen. Hawley’s bill reminds us that DC regulators are gunning for Section 230. To guide regulators through the inevitable craziness, a coalition will soon release a statement offering some guide-rails for those considerations. I will categorically oppose virtually all Section 230 reforms, but I hope we can fruitfully channel the discussions away from the worst reforms. Though we began working on our statement before Sen. Hawley dropped his bill, it won’t surprise you that Sen. Hawley’s “Support for Internet Censorship” bill violates several of the guide-rails we identify.

* * *

Additional Coverage

Thanks to Ashkhen Kazaryan and Sasha Moss for preparing this roundup of coverage, which I’ve supplemented a little:

Academics/Legal Experts

Members of Congress

Civil society

Trade Associations

Media

* * *

Additional Section 230 Readings

Some of my Section 230 writing that might be relevant to this bill:

* Want to Kill Facebook and Google? Preserving Section 230 Is Your Best Hope (2019)
* An Overview of the United States’ Section 230 Internet Immunity (2019)
* Why Section 230 Is Better Than the First Amendment (2019)
* Of Course the First Amendment Protects Google and Facebook (and It’s Not a Close Question) (2018)