Comments on the “SAFE TECH” Act

This year’s tsunami of Section 230 “reform” bills includes the “Safeguarding Against Fraud, Exploitation, Threats, Extremism and Consumer Harms (SAFE TECH) Act” from Sens. Warner, Hirono, and Klobuchar.

This bill proposes over ten different Section 230 reforms, some of which I believe have never been proposed before, and any one of which could disrupt the Internet we know and love. Together, these reforms would reshape the Internet dramatically, in ways that would stun and anger most Americans. Even if you support Section 230 reform, please examine this bill carefully. It’s a radical and clumsy attempt to redress problematic online content, with little concern about the potentially dramatic collateral damage.

What the Bill Says

(The drafters helpfully circulated a redline of the bill against Section 230’s current provisions. Not so helpfully, the redline contains minor errors).

Currently, Section 230 does not apply in four circumstances: IP claims, federal criminal prosecutions, ECPA and related state law claims, and FOSTA-related claims. The bill would add SIX more circumstances where Section 230 would not apply:

  • if a plaintiff seeks an injunction because the service failed to “remove, restrict access to or availability of, or prevent dissemination of material that is likely to cause irreparable harm” (see the drafting oddity note below).
  • claims alleging “discrimination on the basis of any protected class, or conduct that has the effect or consequence of discriminating on the basis of any protected class.”
  • federal or state “antitrust” claims.
  • federal or state claims for “stalking, cyberstalking, harassment, cyberharassment, or intimidation based in whole or in part on sex (including sexual orientation and gender identity), race, color, religion, ancestry, national origin, or physical or mental disability.”
  • claims based on 28 USC 1350 (the Alien Tort Claims Act), which authorizes “civil action[s] by an alien for a tort only, committed in violation of the law of nations or a treaty of the United States.”
  • “wrongful death” actions.

The bill would add three more exclusions to Section 230(c)(1) for any speech that the provider/user:

  • “has accepted payment to make the speech available.”
  • has “in whole or in part, created” [as discussed below, this partially restates the current Section 230(f)(3)].
  • has “in whole or in part…funded the creation of the speech”

The bill also would say that Section 230(c)(1) is an affirmative defense that the provider “has a burden of proving by a preponderance of the evidence” (see drafting oddity below). Finally, the bill would change the word “information” to “speech,” raising the obvious question: when does information NOT qualify as speech? (The bill redline shows this information-to-speech change in two places, but the second redlined change (in Section 230(f)(3)) isn’t in the bill).

Drafting oddity: In two places, the changes only apply to providers of interactive computer services, not users of interactive computer services. But ICS providers are ALWAYS ICS users because everyone online–including all Internet services–necessarily uses an Internet access provider (which the statute defines as an ICS). So one weird possibility is that the existing ICS providers will simply claim Section 230 defenses as ICS users and get around those two changes in the bill.

Why This is a Terrible Bill

I would gladly start my statutory analysis with praise or compliments for the bill. Sadly, once again, I have none. The proposed reforms range from bad to terrible. My top six categories of problems:

1. What Problems Does This Bill Solve? 

The SAFE TECH Act is an omnibus bill. Instead of identifying a specific problem and then crafting a solution carefully designed to fix the problem, the bill tosses a bunch of unrelated reforms into a statutory salad. Omnibus bills are usually terrible, but especially when dealing with something as foundational and nuanced as the rules for online speech. The bill’s disjointed smorgasbord of policy “solutions” is a strong signal that the drafters aren’t sure what problems they hope to solve, if the reforms will solve them, or how Section 230 works.

2. Undermining Section 230’s Structural Integrity.

Currently, it’s virtually impossible for plaintiffs to “creatively” plead their way around Section 230. Section 230 applies to all claims that aren’t in the four statutory exceptions; and the few common law exceptions rarely help plaintiffs because the plaintiffs don’t succeed on the prima facie elements of their claims, and those pleadarounds increase everyone’s litigation costs without changing the result. Section 230 derives much of its utility from how it immunizes against liability for third-party content no matter what claims plaintiffs actually plead.

This bill would add NINE new statutory exceptions to the four existing exceptions. With a menu of 13 workarounds to Section 230 to choose from, future plaintiffs will almost certainly find some exception to Section 230 they can use. If a plaintiff can always plead around Section 230, Section 230 loses most of its effectiveness.

3. Eliminating Section 230’s Procedural Fast Lane.

For reasons I explained in detail here, it matters when UGC services win a Section 230 dismissal. Among other reasons, resolving cases on motion to dismiss substantially reduces the defense costs and removes plaintiffs’ abilities to play hold-up games to capture defense costs. Thus, any reform that disrupts Section 230’s procedural “fast lane” will dramatically change how UGC services behave.

The SAFE TECH Act upends Section 230’s procedural fast lane in at least two ways.

First, the bill says Section 230 is an affirmative defense. That restates the existing law–although a number of courts have said that 230 is an “immunity from suit,” which this language seems to overturn. Nevertheless, courts can and routinely do grant the motion to dismiss when Section 230 is apparent on the complaint’s face. See Ian C. Ballon, 4 E-Commerce and Internet Law 37.05[7] (“Procedural Issues—When to Raise the CDA Defense”).

Could that still happen after the bill? The bill adds that the defendant “has a burden of proving [the defense] by a preponderance of the evidence.” On the surface, this seems to restate the standard evidentiary burdens associated with an affirmative defense. Yet, defendants can never “prove” a defense, using evidence, on a motion to dismiss. So how will courts interpret this language? I think the added language will deter some courts from granting the motion to dismiss, even if the defendant accedes to all of the complaint allegations, because the defendant did not yet satisfy the evidentiary burden. In those cases, the litigation will cost more to reach the same outcome.

Second, even if courts might still grant Section 230-based motions to dismiss, the bill makes that outcome unlikely by making it easy for plaintiffs to allege a fact dispute. In particular, plaintiffs would always allege that the defendant created/funded the creation of the content at issue (whether it’s true or not). That factually unsupported claim should always overcome a motion to dismiss. Cf. Huon v. Denton (a factually unsupported allegation that the defendant ghost-wrote a putatively third-party comment defeated a motion to dismiss).

(Note: Section 230 already excludes defendants who “create or develop” “in part” the content at issue. Courts usually read this language narrowly to mean that the defendant created what made the third-party content tortious/criminal. By incorporating the “creation” exclusion into Section 230’s preconditions, and by adding a new “funded the creation” alternative, courts would likely read those words more broadly than current jurisprudence).

If the bill passes, defendants would rarely win Section 230 defenses on motions to dismiss, even when Section 230 actually applies. The added cost and distraction of litigating beyond a motion to dismiss will lead to many unwanted outcomes, including UGC services removing a lot more non-violative content (proactively to avoid the risk of lawsuits, and in response to unmeritorious litigation demands), litigation trolling and quick settlements for less than the defense costs, and UGC services shutting down speech venues altogether.

4. Pitting Section 230 Against Generating Revenue, When Most UGC Services Need Both.

I believe the drafters wanted to prevent UGC services from profiting from tortious or criminal content/activities. Unfortunately, the resulting language is a textbook example of the Supreme Court’s speech metaphor about burning down the house to roast the pig. The bill’s actual language makes it impossible for most UGC sites to profit from any content, legitimate or not. The provisions will essentially force most services to choose between Section 230 and generating revenue. However, most UGC services need both Section 230 and revenue to survive, so the bill creates a lose-lose Morton’s Fork that would devastate the UGC industry. As Cathy Gellis explains, “this bill would require every Internet service be a self-funded, charitable venture always offered for free.”

The bill would eliminate Section 230 when the service “has accepted payment to make the speech available.” This phrase is the product of sloppy drafting. The bill doesn’t say who makes the disqualifying payment. By definition, every commercial UGC service gets payment from someone. I assume that the bill meant that the payment comes from the content submitter, but that’s not what it says. Furthermore, the bill doesn’t define “payment.” I assume the drafters meant cash payments. If “payment” includes non-cash conferrals of value, then I don’t know what would be left of Section 230. By definition, virtually every UGC submitter values the publication of their content and thus gets “paid” in the form of utility (this is Econ 101–most market exchanges produce surplus for both participants).

4(a): excluding paid ads

Moving on from the lazy drafting, the bill clearly says that publishers couldn’t claim Section 230 for ads they are paid to run. I know this policy approach appeals to many people, but it will disrupt the online ad ecosystem in unwanted ways:

  • it will eliminate all automated self-service ad placement systems. At minimum, those systems will need to add a human into the loop to verify each ad. That will raise compliance costs, some of which will be passed through to increase the costs of advertising.
  • it will create additional risk of liability for publishers, and they will respond by increasing the cost of advertising.

The advertising ecosystem is really struggling during the pandemic, and this bill would compound those woes. Smaller ad-supported UGC services are already having financial difficulty, and the increased costs of running an ad-based business–plus the loss of advertisers who can’t afford the higher prices–will put publishers out of business or drive them to alternative revenue streams, such as paywalls. Eliminating the small ad-supported publishers from the industry will deepen the competitive moat of the existing incumbents–Google and Facebook–who will appreciate Congress’ help cleaning out their competition.

4(b): excluding online marketplaces

In addition to removing traditional display ads from Section 230, the bill apparently targets all paid listings in online marketplaces, including Amazon’s marketplace, eBay, Airbnb, and many more. (I don’t know how the statutory language would apply to marketplaces that accept free listings but charge a transaction fee on the deals; I think the HomeAway case may have eliminated Section 230 in that circumstance anyways). I don’t know what marketplaces will be able to survive without Section 230 protection, but the collapse of that niche would really bum out Americans. Without the competition provided by online marketplaces, prices will go up; many small businesses that rely on marketplaces will fail–and so will the jobs they create; and consumers will spend more time and money on search costs.

So far, I’ve shown how the bill would hurt or eliminate small ad-supported publishers and many online marketplaces. Now, I’ll show how the bill would hurt subscription business models too.

4(c): excluding subscription services

The bill says that Section 230 wouldn’t apply to any speech that the UGC service “has accepted payment to make” available–which seems to include any editorial content where a service charges the user to host or publish the content. That includes every web host that charges subscription fees–WordPress, Flickr, Dropbox, Google Drive (for overage usage), even B2B services like AWS. Also, any paywalled gardens would apparently lose Section 230 protection for any UGC components within the walled garden, such as message boards, chats, or comments.

Ironically, the bill gets things backwards. People who pay for hosting services are less likely to engage in anti-social behavior than those using the free hosting resources, but the bill counterproductively privileges the freely hosted content. It also puts stress on all freemium hosting business models, where the -mium part comes with greater liability risk and thus either can’t work or will drive prices up and make those tools less affordable to already marginalized voices.

4(d): Implications

Now you can see why this bill will ravage the UGC ecosystem. It hammers ad-supported and subscription UGC services and online marketplaces. What business models are left for UGC services, how profitable will they be, and how many players can afford to succeed? Ironically, the bill’s concerns about money don’t appear to reach pay-to-remove services (such as the mugshot sites or the old revenge porn sites), which are often the most troubling sites.

These business model changes seem destined to drive more conversations to private messaging services like email and instant messaging. Is that a net win? Those conversations become less visible to plaintiffs and law enforcement. Also, I’m not opining on how this bill would affect Section 230’s availability to email service providers and other private messaging servicing providers, though I’d love to hear other experts’ thoughts on that question.

5. Pitting Section 230 Against Compensation for Content Producers.

In its pig-roasting effort to keep tortious/criminal content from getting financially rewarded, the bill also hammers UGC acquisition.

The bill says that Section 230 doesn’t apply when a defendant “funded the creation of the speech” (in part or whole). This is also lazy drafting. It’s not clear how we’ll know the but-for/proximate causation of when money spurs the production of UGC. Also, if the UGC service is being sued, what evidence will it have about the motivations of the creator?

I think the most likely reading is that if a UGC service pays a user for supplying content (through license fees, revenue shares, or some other mechanism), the UGC won’t qualify for Section 230. If so, the bill would clearly overturn cases like Blumenthal v. Drudge, where AOL avoided liability for carrying Matt Drudge’s content even though AOL paid a substantial license fee to Drudge (and was his main source of income). Other entities that would likely lose Section 230 protection for third-party content:

  • self-publishing services where the content suppliers set the price, like Amazon Createspace, Kindle, and the Gumroad service I use to self-publish my casebooks
  • ad networks that run ads on third-party services, such Google AdSense.
  • services that pay license fees/revenue shares to spur new creative productions outside of the traditional studio/label/intermediary model, including YouTube, TikTok, Patreon, OnlyFans, Substack, and many others.

The bill would motivate services to publish only two types of content: (1) UGC that users are willing to supply for free, which counterproductively will often be lower-quality and has a higher risk of being anti-social, or (2) professionally produced content, where the online publishers will manage their legal risks through pre-vetting (because there is so much less content), indemnities/insurance, and the producers’ own fears of their liability. Even if UGC services still accept UGC without paying for it, they will have few options to monetize it; and if they abandon UGC, their content licensing costs will go up in ways they may not afford. So who will be left standing after this content monetization/acquisition double-whammy? And how many small self-employed content producers can survive the downdraft? (Hasta la vista, YouTubers).

6. A Closer Look at the Exceptions

Some of the new proposed exceptions are unclear in scope. Every ambiguity in every exclusion represents lots of potential litigation and uncertainty to defendants.

Antitrust. What is an “antitrust” claim? Would it include any allegation that the defendant had anti-competitive animus? What about “unfair competition” claims? That includes a huge swath of behavior.

Wrongful Death. Does this apply only to claims labeled as “wrongful” death, or would Section 230 categorically not apply to any claim where a person died? Note that the cases against social media for allegedly materially supporting terrorists failed on a range of grounds, including lack of proximate causation, statutory grounds, and Constitutional limits. Section 230 reform won’t change those outcomes.

Alien Tort Claims. 28 USC 1350 only applies to “aliens.” I’m not an expert in this law, but it seems like the bill would categorically exclude Section 230 for foreigners’ claims. If so, it seems odd to give foreigners greater legal claims than US residents. Plaintiffs’ lawyers would get around Section 230 by adding a foreign plaintiff where possible.

Also, how do the drafters think UGC services will manage this risk of alien claims? UGC services can’t easily implement content moderation strategies that identify and eliminate the legal risks only to foreigners. Instead, UGC services would build content moderation tactics that presume any legal risk on their site would fall into this exception. While Section 230 would still possibly help once in court, this exception would cause UGC services to prospectively moderate content as if Section 230 doesn’t exist at all.

Discrimination. The discrimination exclusion is nonsensical because content privileging is inevitable. Every publication decision represents a form of bias. Furthermore, this exclusion will surely backfire. People with majority attributes would LOVE to weaponize anti-discrimination laws to preserve their power. We’ve already seen numerous #MAGA lawsuits to this effect (e.g., Lewis v. Google, Wilson v. Twitter, Domen v. Vimeo)–despite Section 230 currently applying to those claims. Assume that the paradigmatic discrimination plaintiff would be a Christian heterosexual white male suing because the UGC service suppressed their misogynistic, anti-Semitic, racist, or other hateful content (allegedly) due to their “protected” majority attributes. Is that really what the drafters want?

Harassment/Stalking/Intimidation. I’m not sure what constitutes an “intimidation” claim. I think many “creative” plaintiffs will be able to claim that they suffered intimidation, harassing, or stalking from a wide range of anti-social behavior online, so what isn’t part of this exception? Also, due to sloppy drafting, the exclusion limits claims to those “based in whole or in part on sex (including sexual orientation and gender identity), race, color, religion, ancestry, national origin, or physical or mental disability,” but doesn’t make clear if that modifies only the intimidation claims or extends back to limit all harassment and stalking claims.

Injunction. I previously blogged the implications of excluding injunctions from Section 230 when I blogged the PACT Act. That discussion ran over 1,000 words, so I won’t recap all of that here. I’ll just point out that the exclusion raises the question of what UGC services would do in response to the pending litigation. Would they show up in court, at some expense, to defend their interests, or would they ignore it and just wait for the injunction to roll it? What do the drafters want to happen, and what legal mechanisms will allow that to happen?

One Final Thought. The laundry list of exclusions raises a broader question: how do we know these exclusions are the right ones? In other words, how did the drafters decide which victims’ communities to privilege and which, by implication, not to privilege? I’d love to hear more about the analysis that went into these privileging decisions and the decision not to include other victims’ exceptions. More importantly, assume that post-passage every other victim’s group will demand a Section 230 exclusion too. What intellectually defensible rationale will Congress use to decide if they are worthy? The lessons from the FOSTA exclusion might help here.


As I’ve complained before, I can’t wrap my head around how Senators and their staffers are working hard to tear down Section 230, when at the same time, (1) we have must-fix problems like the pandemic, the aftermath of a coup, the doubled-down efforts to lock in minority rule through voter suppression, economic suffering, racial inequality, and so much more, and (2) the UGC services have been the lifeline bridging us through the pandemic and helping us stay safe.

If Congress does think this is what needs addressing today, we need to have a candid conversation about the overall value of UGC. I care about it passionately, but many people (including Congressmembers) apparently don’t care if UGC disappears. If you support this bill, you probably feel that way too.

The bill’s press release claimed the bill would make UGC services take their content moderation responsibilities more seriously. This is hubris. UGC services will not interpret this bill as a motivation to step up; the higher legal risk/costs and decreased monetization potential will tell them to GTFO of UGC altogether. The net effect will be that some online publishers will migrate to walled gardens of professionally produced content; others will fold altogether; and only Google and Facebook might survive the UGC cataclysm, though they would curtail many current tools for user expression because they can’t control the associated legal risk.

As a result, this bill will wipe out hundreds of thousands of jobs and hundreds of billions (maybe trillions) of current market cap. More importantly, it will eliminate most of the things we love most about the Internet. That’s why I think this bill would radically reshape the Internet and make our lives worse.

For more stinging criticisms of the bill, see Mike Masnick’s writeup.

REMINDER: In July 2019, a huge coalition put together principles for evaluating reforms to intermediary liability rules. They would be helpful for drafters of legislation like the SAFE TECH Act. Not surprisingly, the SAFE TECH Act deviates from several of the principles (including #1, 2, 3, 5, and 6).

Prior Blog Posts on the 117th Congress’ Efforts to Destroy Section 230