On October 21, a New Mexico judge ordered Meta to pay the state $25 billion for violating its Unfair Practices Act. That penalty is massive. But the bigger problem is how the judgment treats the gap between a platform’s moderation promises and its imperfect enforcement as deception, penalizing platforms for moderating protected speech and sidestepping Section 230 and the First Amendment. If nothing is done to stop these attacks on free expression, we will all lose access to online platforms, big and small, as we know them.
New Mexico Is Effectively Enforcing Speech Codes
So what did Meta do that the court said warranted billions in fines? While most of the case concerned privacy issues related to Cambridge Analytica, another major basis was that Meta had supposedly deceived users by allowing misinformation, hate speech, and violent content on its platforms after claiming it would take down such content.
These are issues I know firsthand because I was a subject matter expert on hate speech and on violence and incitement for what was then called Meta’s Organic Content Policy team. My team wrote the policies that Meta’s enforcement systems then applied.
The verdict found various statements by Meta to be deceptive. For example, the jury found these statements by Mark Zuckerberg deceptive:
“We do not profit from misinformation or hate, and we do not want this content on our platforms.”
“Well, if we see harmful misinformation on the platform, then we take it down. It’s against our policy.… Now, do we catch everything? Of course there are mistakes that we make or areas where we need to improve.”
“There are really two core principles at play here. There’s giving people a voice so that people can express their opinions. Then, there’s keeping the community safe, which I think is really important. We’re not going to let people plan violence or attack each other or do bad things.”
Other official Facebook communications and its community standards make similar statements.
All of these statements attempt to set a principle that governed Meta’s content moderation decisions (though you could find similar statements and policies on nearly every platform or forum). Meta’s hate speech policies covered a range of insulting and offensive speech aimed at people because of their protected characteristics.
But crafting and enforcing content policies is quite the challenge. Tech writer Mike Masnick has argued that “content moderation at scale is impossible to do well. More specifically, it will always end up frustrating very large segments of the population and will always fail to accurately represent the ‘proper’ level of moderation of anyone.” Two problems explain why: scale and disagreement.
The first problem is scale. Each policy statement made by Meta reflected what it wanted to remove from its platform. Platforms enforce these policies with automated tools and human reviewers across billions of posts, and both make mistakes. Enforce aggressively, and more nonviolating speech gets removed; enforce cautiously, and more violating speech stays up. There is no option to avoid errors, especially across millions of pieces of content every day, which is the point of Zuckerberg’s caveat about not catching everything.
The second problem is disagreement among different viewpoints. People disagree about what counts as hate speech, violent speech, or misinformation. For example, Meta’s policies did not allow users to attack people defined by protected characteristics but generally allowed criticism targeting ideas, concepts, or institutions associated with those characteristics. So you could not say “I hate Christians,” but you could say “I hate Christianity.” You could not say “trans people are evil,” but you could say that “transgenderism and transgender-related surgeries are evil.” Many people disagree with that and other lines of the policy.
This exemplifies one of the major problems with hate speech as a concept—it is inherently subjective. The same is true for many other concepts, like misinformation, which is why the First Amendment prevents the government from censoring speech that some may consider hateful or misleading. And while the First Amendment protects the right of private institutions, both offline and online, to set their own standards for the speech they allow, elevate, restrict, or prohibit within their services, that same subjectivity means different groups will always object to wherever a platform draws the line.
The New Mexico judgment, then, is an exercise in impossibility, trying to make platforms remove hate speech, misinformation, and any other violating content perfectly. Yes, internal documents show that Meta’s leaders knew enforcement fell short of the company’s public principles—as every leader of every platform knows it does. But no platform can enforce perfectly at scale, and a company that says it wants to remove offensive, harmful, or objectionable content is not promising that it has.
New Mexico Treats Extra Review and Disagreement as Liability
New Mexico also argued that Meta lied about giving special treatment to public figures. The court verdict cites Facebook statements such as:
“We want to make clear that we remove content from Facebook, no matter who posts it, when it violates our standards. There are no special protections for any group—whether on the right or the left. ‘Cross Check’ … simply means that some content from certain Pages or Profiles is given a second layer of review to make sure we’ve applied our policies correctly.”
“I know many of you think we should have labeled the President’s posts in some way last week. Our current policy is that if content is actually inciting violence, then the right mitigation is to take that content down—not let people continue seeing it behind a flag. There is no exception to this policy for politicians or newsworthiness.”
The first quote refers to Meta’s “Cross Check” or “X Check” system, which gave certain high-profile users an extra review before their content was removed. Imagine an automated system erroneously removing a post by Donald Trump or Joe Biden about a new government policy or political endorsement. Even a system that is right 99.9 percent of the time makes lots of errors at scale, and accusations of bias and election interference would follow immediately. Similarly, imagine a human rights activist in an unfree and dangerous part of the world having their speech reported at every turn by those seeking to silence them. Inevitably, some of their content will be wrongly removed. So platforms may add a second look for such accounts. Such extra review does add delays before violating content is removed, but that is a trade-off when trying to avoid consequential mistakes.
The second quote refers to President Trump’s post “When the looting starts, the shooting starts” during riots that followed Black Lives Matter protests. Many people wanted the content taken down for inciting violence, and there was intense internal debate about how to handle it. I was not directly involved in that escalation, but my view then and now was that it did not violate Meta’s policies because it could reasonably be read as describing natural consequences—that is, when looting starts, violence follows. Other platforms like Twitter said it violated their rules. The point is that reasonable people and platforms disagree about how to moderate such content.
New Mexico’s attorney general and its courts want to second-guess these decisions. New Mexico argued, and a jury found, that Meta’s statements about treating everyone equally were deceptive. Allowing partisans to prosecute platforms for content moderation decisions they don’t like is dangerous. Blue states could just as easily charge that promises to remove hate speech were false because certain conservative views on charged cultural issues remained up, using arguments not so different than those made by New Mexico. Red states or conservative plaintiffs could find platform statements promising to let users express themselves and then frame moderation of conservative speech as a deceptive practice. Groups on the right, like Prager University, have tried this so far without success.
Section 230 and the First Amendment Should Have Mattered
The New Mexico judgment, if allowed to stand, would radically shift the content moderation promises and practices of countless platforms. Any aspirational moderation statements or clear content policies would create ruinous legal risks. If other states follow this template, attorneys general and other plaintiffs could sue platforms over any content moderation decision they don’t like by alleging it is unfair and deceptive based on general statements about moderation. Platforms can be liable if they moderate too little, too much, too quickly, or too slowly. This is a no-win moderator’s dilemma, and the rational response will be for platforms to promise less, be less transparent, and moderate their users’ speech in ways that minimize liability.
Congress solved this problem when it enacted Section 230, which says, “No provider or user of an interactive computer service shall be treated as the publisher or speaker of any information provided by another information content provider.” In other words, platforms are not liable for content created by internet users. It also added that no platform could be held liable for “any action voluntarily taken in good faith to restrict access to or availability of material,” that is, to moderate content. These few words created the internet as we know it today because they left liability with those who posted the content, not the platforms themselves.
It is surprising, then, that the judgment does not even mention Section 230. If it had, the court would have needed to consider various precedents to determine whether Section 230 applies. When considering how Section 230 applies to a given case, courts ask whether the claim against a platform treats it as a publisher or rests on something independent, like a specific promise. In Barnes v. Yahoo!, Inc., the Ninth Circuit let a claim proceed against a platform that allegedly breached a direct promise to plaintiff Barnes to remove a harmful profile. Barnes claimed she relied on that promise and took no further action, and the profiles stayed up for about two more months. Recent Ninth Circuit decisions have concerningly stretched that exception to a platform’s terms of service (Calise v. Meta) and its promises to ban abusive users (Estate of Bride v. Yolo), and New Mexico would no doubt cite them.
But even those harmful Ninth Circuit precedents did not involve liability for general, aspirational statements by executives in interviews, earnings calls, and newsroom posts about how a platform tries to moderate. Barnes itself said a vague promise of a “general monitoring policy” does not create contract liability. But the New Mexico case flips this on its head, turning every content policy or statement about content policy into a legal promise, and the exception swallows the rule. California’s Court of Appeal held in Prager University v. Google that YouTube’s generalized public statements about its moderation did not create a contractual obligation, and that Section 230 barred claims aimed at its moderation decisions.
The judgment also does not address the First Amendment. In effect, New Mexico is using its courts to demand specific kinds of moderation, such as the rapid removal of content without additional review. But Florida passed a law that would have demanded the opposite. Florida’s SB 7072 would have barred platforms from removing, labeling, or deprioritizing content from journalistic enterprises or posts by or about political candidates. In the 2024 NetChoice cases, the Supreme Court signaled that Florida could not require certain content moderation practices without violating platforms’ First Amendment rights, and that principle should apply equally here.
Liability Cripples Online Speech
Recasting fallible content moderation or content policies the state disagrees with as a form of deception is, then, an attempt to sidestep the protections of Section 230 and the First Amendment with major implications for online speech.
This is not hyperbole, as the size of the New Mexico judgment should make clear how serious the risk of liability is. Even for a large company like Meta, the prospect that each state could sue for tens or hundreds of billions of dollars for past content moderation decisions, as well as liability for future decisions, will force it to drastically change the services it offers, restrict speech on any politically sensitive topic, and aggressively remove constitutionally protected speech or else be driven out of business or out of these states.
Many smaller platforms similarly allow user-generated content and publish their content rules—rules that could be turned against them, creating existential liability. Websites you wouldn’t even think about, like Wikipedia, the New York Times comments section, Google reviews, Yelp ratings, and many more forms of moderated user content, would instantly become legally untenable.
The consequences of this decision may be felt first by Meta, but if it stands, it could change online speech for everyone. Platforms that promise to moderate and fall short, as every platform will, now face billion-dollar penalties. The safest response is to promise less and take down more. That is how your speech pays for New Mexico’s penalty.