Constitutionality of California’s Child Online Safety Law Disputed in Court of Appeals

On Wednesday, the US Ninth Circuit Court of Appeals heard oral arguments for a legal challenge brought by NetChoice, a tech trade association, to a California law that takes a ‘safety by design’ approach to protecting children online. The California Age Appropriate Design Code (CAADCA) was originally challenged on First Amendment grounds by NetChoice in late 2022 (NetChoice v. Bonta), and last year, a district court judge issued a preliminary injunction blocking the law. Now, California is challenging the enjoinment, and the appeals court is tasked with assessing the constitutionality of certain provisions of the law and on what basis they should be analyzed.

One of the requirements of the CAADCA is that online businesses likely to be accessed by children must conduct a Data Protection Impact Assessment (DPIA). These are meant to “assess and mitigate risks” to children that arise from a business’ data management practices and whether the design of its services, products, or features could lead to children accessing potentially harmful content, among other requirements.

While the state argued in the lower court that the law “merely regulates business practices regarding the collection and use of children’s data, so that its restrictions are only of nonexpressive conduct that is not entitled to First Amendment protection,” the district court judge did not agree. Instead, the decision determined that the provisions of the DPIA were “regulat[ing] the distribution of speech and therefore trigger First Amendment scrutiny.”

During Wednesday’s hearing, Judge Milan D. Smith, Jr., in a tense exchange, seemed to agree with the district court, arguing that making a determination on whether data is harmful to children inherently implicates content moderation. “If you get there, then you’ve got a First Amendment challenge, right?” However, California’s Deputy Attorney General and counsel, Kristin Liska, maintained that the law is meant to look at “the use of data [that] potentially exposes children to harmful content, not into an analysis of an exhaustive array of what content is harmful or not.”

The Court’s questioning was very much in line with NetChoice’s arguments against the DPIA. "Today's hearing ripped the mask off California's AADC, further exposing the state’s attempt to censor the internet,” said NetChoice Litigation Center Director, Chris Marchese, in an emailed statement to Tech Policy Press. Marchese, who was in the courtroom on Wednesday, argued that the Court hammered home “the law's threat to online speech and access to information,” which he characterized as a “muzzle on online expression” rather than a privacy protection.

“The courts seem to be very focused on the fact that they want to basically rule that the DPIA is unconstitutional,” said Gaia Bernstein, professor of law at Seton Hall Law School, in a press briefing organized by the Tech Justice Law Project following oral arguments. “For me, what matters is how will they frame it?” If the bench ties its decision to 'harmful content,' Bernstein said, it won’t be as bad for efforts to protect kids online. On the flip side, if the court doesn’t explain its decision, it “could be viewed as a precedent that basically shuts down efforts to regulate design.”

Megan Iorio, senior counsel and amicus director at Electronic Privacy Information Center, sees this likely interpretation as “good news” for content-neutral data and design regulations generally. “The judges otherwise seem to disagree with the lower court's dangerous decision that every privacy law is presumptively unconstitutional,” she said in the press briefing. Instead, the survival of the non-DPIA provisions will “likely depend on whether the panel thinks the DPIA provision is severable from the rest of the law.”

Earlier this month, the US Supreme Court unanimously returned two cases (Moody v. NetChoice and NetChoice v. Paxton) to lower courts regarding Florida and Texas laws that regulate large social media platforms. The opinion found that the laws were written too broadly, covering more than just platforms like Facebook and YouTube, and that neither of the lower courts had “address[ed] the full range of activities the laws cover, and measure the constitutional against the unconstitutional applications.” The circuit courts must now first assess the scope of the Florida and Texas laws, and then decide which of the laws’ applications would violate the First Amendment.

Although the Supreme Court’s Moody opinion came after the district court’s opinion on the CAADCA, NetChoice Counsel Robert Corn-Revere argued on Wednesday that Moody didn’t substantially change the law underlying the Bonta suit. Rather, it “affirmed the analysis” that if there are “enough unconstitutional provisions of a law,” it can be challenged facially, rather than provision-by-provision.

The Ninth Circuit panel appeared to challenge this framing. Judge Anthony D. Johnstone said the courts must distinguish between severability on the one hand, and a facial challenge on the other, as one is state law and the other is constitutional law, respectively. He understood Moody as “slicing and dicing for the purpose of a facial challenge,” and wondered whether the district courts should scrutinize the CAADCA section-by-section before moving onto a severability analysis. In California, if certain provisions of a law are found unconstitutional, severability allows the remaining provisions to stay in effect, but the courts often make the determination of whether to void the entire measure or not.

Judge Bennett suggested that this approach punishes the California legislature for sticking “different provisions that deal with some different things” in the same bill. He asked NetChoice’s Counsel whether it was their view that the courts should look at the provisions altogether rather than examining them as distinct statutory provisions, even if some sections would pass constitutional scrutiny. NetChoice maintained that the courts should examine the law as a whole when determining how much speech is being regulated by it. “The answer is a vast amount of speech,” Corn-Revere said.

Liska seized on this splinter in her rebuttal: "The way to view severability is if you were to take an entire piece of paper and you were to carve out a hole from the middle, which is the part that you view as unconstitutional, can the rest of the piece of paper stand?” She argued that the CAADCA must be examined in its entirety for a severability analysis that drills down at each subdivision in order to determine the constitutionality of each.

“The hearings represented a real confusion, perhaps, with the various standards that are at play, and also an uncertainty about how to apply the Supreme Court's recent precedent in the NetChoice v. Moody and Paxton cases,” Meetali Jain, director of the Tech Justice Law Project, said at the press briefing. “But certainly, it’s a very critical moment for looking at tech regulation” and whether it can “really withstand constitutional scrutiny per the First Amendment in this country.”

Below is a lightly edited transcript of the oral arguments for NetChoice v. Bonta. Please refer to the official video when quoting.

Clerk:

All rise...The United States Court of Appeals for the Ninth Circuit is now in session...Please be seated.

Judge M. Smith, Jr.:

Good morning, ladies and gentlemen. Welcome to the Ninth Circuit. We have three cases for argument this morning.

[...]

Judge M. Smith, Jr.:

We will now hear arguing the case of NetChoice versus Bonta. All right, Ms. Liska, I believe you're up first.

Kristin Liska:

Good morning, your honors. Kristin Liska on behalf of the Attorney General and I'd like to reserve about five minutes for rebuttal. I'll try to keep an eye on the clock. May it please the court, the California legislature unanimously enacted the California Age Appropriate Design Code Act to protect children and their privacy online. The legislature was motivated by documented harms to children from online companies, data and privacy practices, including addiction-like behaviors, lack of sleep, negative impact on educational attainment, exploitation, and detrimental mental health outcomes.

Judge M. Smith, Jr.:

Counsel, I think all of us would say the objective is one that we could only laud, but we are dealing with the First Amendment here and we have some questions about the nature of what is included in this bill. I'm going to use an abbreviation, but I want to make certain the record is clear, I'm talking about the Data Protection Impact Assessment, which is at Civil Code 1798.99.33A. So, I'm going to call it the DPIA, okay? This law forces the private parties to opine on what they think is a material detrimental to children. How isn't that compelled speech?

Kristin Liska:

So, if we're talking about just the DPIA report itself, I don't think that we would dispute that the requirement to prepare a report can be considered compelled speech.

Judge M. Smith, Jr.:

Okay and if it is compelled speech, don't you run into a First Amendment problem?

Kristin Liska:

We would argue that it's constitutional as a disclosure, since it's a factual non-controversial disclosure.

Judge Anthony D. Johnstone:

Well, this is framed as data management practices, but isn't the first of the data management practices that are called on for these companies, whether the content of what is provided online is potentially harmful? How is that not a content-based restriction rather than a pure factual disclosure?

Kristin Liska:

I think that what's critical is that the purpose of the DPIA requirement is to look at how the data practices pose risks and material detriment to children. And so in the process-

Judge Anthony D. Johnstone:

Do the data practices include an evaluation of whether the company's online content is harmful to children?

Kristin Liska:

... no. The data practices would not.

Judge Anthony D. Johnstone:

... content is harmful to children?

Kristin Liska:

No, the data practices would not. Our understanding of the statute is that the DPIA requirement would look to, “How does the use of data potentially expose children to harmful content?” not into an analysis of an exhaustive array of what content is harmful or not. And I can provide an example. For instance, a company might use data for targeted or behavioral advertisements where they take the personal information they've collected to try to determine which advertisements a user is more likely to click on. So perhaps they then would use advertisements that might advertise products that are illegal to sell to minors, tobacco or alcohol products. Another example might be a company may use data to determine what order to display posts in. A post that has more likes-

Judge Anthony D. Johnstone:

Okay. Those seem to be addressed by Romanette 6, 7, and 8, or perhaps 7 and 8. But there's also Romanette 1, which speaks to content, 2, which speaks to contact, 3 to whether it permits children to witness harmful conduct, which again seems to be content. What are examples of those that you think would not trigger content-based scrutiny under the First Amendment?

Kristin Liska:

So again, I think that the focus of this analysis is not on the content itself. There's no requirement to look through and decide what type of content is harmful. The question is about the data use. So an example may well be if you use prior search history, for instance, so say a minor child does a research project on eating disorders for a health class, and then that search data is put into an algorithm, and then videos that promote eating disorders are recommended to the minor. That's an example of how data practices might lead to minors viewing harmful content.

Judge M. Smith, Jr.:

But it seems to me that you require these companies to make a decision, to have an opinion about the nature of what's involved there. As you know, a lot of the amicus briefs and so on talked about things that can disturb children generally, like climate change or wars or even internally local violence and so on and so on. But you're asking these companies to have an opinion about these things and express it, and it's not like a cigarette thing. There's so much tar in this and so on. It's not a commercial thing. You're asking them to make a determination, and that's what I'm troubled about. It seems to me that when you are asking them to make that, it's compelled speech and you do bring in the First Amendment and probably at a strict scrutiny level. What am I missing?

Kristin Liska:

So a couple of responses to that, Your Honor. I think, first, the DPIA requirements certainly do not require companies to make an exhaustive list of what content is harmful. And perhaps it might be helpful to–

Judge M. Smith, Jr.:

Well, maybe not an exhaustive list, but they have to state with respect to any particular app how it would be harmful to children. They have to consider these things, right?

Kristin Liska:

They need to consider the data use. We would not argue that they need to look and decide what content is harmful. And it may be helpful if we look at, there is a sample of a DPIA in the record, and this is on, I don't have the volume number in front of me, but ER551 has an example of a DPIA that does this analysis. This is from the UK, but the law is clear-

Judge Mark W. Bennett:

Counsel, going to Judge Smith's question, I'm looking for example at two little eyes, whether the design of the online product, service or feature could lead to children experiencing or being targeted by harmful or potentially harmful contacts on the online product, service or feature. And when I read that as an example, it strikes me as doing exactly what Judge Smith described it as doing as opposed to some high-level data analysis.

Kristin Liska:

Again, I think our view of how the statute operates is that there's no requirement. I think everyone here would agree that there is harmful content that exists in the world for children and there are harmful contacts for children, people who seek to exploit them. I don't think anyone disputes that fact. There may be a dispute about what falls within there, but we all-

Judge M. Smith, Jr.:

But are you confessing, if you will, that the state is trying to regulate the content?

Kristin Liska:

No, Your Honor, we're concerned about the data collection practices. So if we-

Judge M. Smith, Jr.:

But seriously, I mean come on. If data's just a report of numbers and what it says, you're requiring them to make a determination whether what they have there is harmful to children. It's not just numbers. They have to decide whether it is and on some basis. I'm looking for you to correct me if I'm wrong. To me, that's a requirement to moderate or admit content. And if you get there, then you've got a First Amendment challenge, right?

Kristin Liska:

Again, we disagree that that's how the statute operates, Your Honor. If you look at the language, it says to identify the use of the purpose of the service, how it uses children's information, and the risks of material detriment to children that arise from the data management practices.

Judge M. Smith, Jr.:

But they only need to do that with respect to harmful content. Who decides what's harmful?

Kristin Liska:

So again, if we look at it this way, and again, I think maybe looking at the example with the record of what a DPIA looks like-

Judge Anthony D. Johnstone:

Well, an example from a jurisdiction that does not have a First Amendment is probably less helpful for us.

Kristin Liska:

No, but I think the point that I'm attempting to make, Your Honor, is that in the world in which we all agree there's harmful content out there, one can analyze the extent to which such content could be exposed to minors by data practices without having to get into what is that whole universe of content. I don't think anyone disputes harmful content is out there. And so the question isn't, what is that universe of harmful content we're concerned about? What is that world of things we don't want, we're worried about might hurt children? It's how does the way you use their data potentially expose them to that universe, which we know is not an all set.

Judge Anthony D. Johnstone:

What happens on the margin if there's a disagreement about whether the triggering content is harmful? So, for example, Judge Smith's examples and the examples in the briefs about images of war or climate change or things that do seem to have harmful effects on children but are also true and protected by the First Amendment.

Kristin Liska:

The question is simply the extent to which this harmful content, which again, it's not a universe that we need to define, whether the data practices might expose children to it. But I do want to take a further step back and just note that there may be some of these difficult questions with the margins, but this is a facial challenge that was brought by plaintiffs. They're abundantly clear from, I think it's paragraph four of their complaint, that this is a facial challenge. And Moody is clear that when analyzing a facial challenge, the question is, does the unconstitutional application substantially outweigh the constitutional. We're looking at–

Judge M. Smith, Jr.:

Well, it's true what Moody said in that regard, but if we're just limited to the DPIA, is it your argument that we cannot analyze that portion of this legislation on a facial basis?

Kristin Liska:

We would say that you can, Your Honor, but there's a whole swathe of things in the DPIA that are not at all related to content. For instance, harmful contacts could refer to situations where, ‘do they allow any adult to message a child online without first having parental permission or without allowing some prior connection?’

Judge Mark W. Bennett:

Counsel, you've certainly argued that the statute is severable.

Kristin Liska:

Yes, Your Honor.

Judge Mark W. Bennett:

Did you argue that a facial challenge is impermissible here other than because there's nothing wrong with the statute? I mean, obviously you didn't cite Moody in your briefs because it hadn't happened yet, but did you argue that only an applied challenge could work here, that a facial challenge is a non-starter?

Kristin Liska:

We may not have argued it in those terms, but I think Moody itself is illustrative on this point. The parties in Moody were very focused on one application of the statutes and the court in Moody, and I believe Justice Alito says this quite clearly in his concurrence, the court in Moody nonetheless said that it is the court's duty to apply the proper standard regardless of perhaps how the parties have framed their arguments. And since this is a facial case and has always been, we think that at the very least a reevaluation of this in light of Moody's discussion of how to do a facial challenge is appropriate.

Judge Anthony D. Johnstone:

There's kind of layers to this statute. We've got the DPIA, I think we've talked a little bit about the subsections of 1A, the Romanette, the content, conduct, and other personal information. At what level should we be evaluating the facial challenge for purposes of what's at stake that would then lead us into a severability analysis of what's left?

Kristin Liska:

I think at the very least, the court needs to look at each specific individual subsection of regulation and whether or not how that specific subsection operates. We think that's one of the fundamental flaws of the lower court's decision. So, for instance, the DPIA requirement is distinct from the subsections that involve no collecting geolocation data or requirement to have a mechanism to address privacy concerns. So at the very least, each of those subsections should have been addressed individually as to whether it even regulates speech to begin with, and then what is the proper First Amendment challenge for that specific subsection.

Judge Anthony D. Johnstone:

What about the sub subsections of the DPIA? My take on it is it seems like the parties as well as the district court have dealt with section 31.1B and all of those eight subsections as a batch. Is there anything in the briefing or the record that would allow us to distinguish, for example, between the sub Romanette 8, discussion of personal information or discussion of advertising versus the calling out of content in Romanette 1?

Kristin Liska:

We would think that the court is perfectly capable of doing that. I believe one of our, it may have been in a footnote notes, that is the sort of granularity that the lower court failed to do with severability analysis. Since we have argued that the lower court should have taken a much more nuanced and granular approach to severability and looked at the specific sections it thought were unconstitutional vis-a-vis the remainder–

Judge Anthony D. Johnstone:

So the enforcement provision, and this is the discussion in terms of the severability, the enforcement provision, a condition precedent for enforcement is whether a DPIA is on file. If assuming for purposes of argument, the DPIA requirement is unconstitutional, what's left given that any remaining enforcement is keyed to the DPIA?

Kristin Liska:

So the DPIA only operates as a notice and cure to the extent that it's been completed. So were the court to declare to enjoin the DPIA requirement, that would enjoin the notice and cure safe harbor, but the remainder of the law would remain enforceable because the rest of the enforcement provisions would still be intact. Companies would just lack that safe harbor noticing cure provision that flows from preparing the DPIA.

Judge M. Smith, Jr.:

But Counsel, I'm a Californian, my colleagues are not. So I followed this as it went along and it seemed to me that it was very clear that the 90-day safe harbor was a negotiated provision primarily with the industry that had to relent to this. It was a safe harbor. The idea was to get people to voluntarily cooperate. It seemed to me, based on the record, that the 90-day consultation period was indeed a condition precedent to enforcement. What's your best argument that that is not the case?

Kristin Liska:

Two responses to that, Your Honor. The first is that I think if you apply the severability test under California law that the grammatical, operational and volitional severability, all three of those criteria are met. You can excise it without making the statute unreadable. The statute can operate just fine without a 90-day notice and cure, though I'm sure–

Judge M. Smith, Jr.:

Well, let me put it this way, arguendo, if the 90-day provision is a condition precedent that is essential to enforce, which my colleague has been talking about, how can it then be severable?

Kristin Liska:

Your Honor, my second response to this is that this court has brought equitable discretion when it comes to fashioning an injunction and should this court feel that the only way to preserve the remainder of the statute, because we think the legislature would've wanted to keep as much as possible, especially when we're talking about limitations on data use and privacy settings. That was really what the legislature worried a lot about. It can certainly fashion an injunction that retains the 90-day notice and cure, but just severs the DPIA requirement.

Judge M. Smith, Jr.:

Let me ask you this, and this is all arguendo, okay? If the California legislature knew that the Ninth Circuit Court of Appeals was going to knock out the DPIA, would they have gone forward with the rest of the legislation without any report, without any condition precedent to enforcement?

Kristin Liska:

I think so, and I think that that flows from both the factual findings that we see in the statute itself and from the legislative record, which shows a lot of concern about children's privacy and harm stemming from data use. And while the DPIA is part of how the hope was to remedy those, the remaining substantive provisions of the act are also important in helping with that. I see I'm running low on time–

Judge Anthony D. Johnstone:

Could I just, coming back to something you said, Ms. Liska, about how the notice and cure could be saved, how is that even grammatically possible looking at C1 where the thing that is noticed is if they have complied with the DPIA? I just want to be clear that I understand how you see this working in a world without the DPIA.

Kristin Liska:

I think that the way that you could do it is if you look at the language of C1, which starts with if a business is in substantial compliance, you could just excise that preparatory clause and start with the Attorney General shall provide written notice to the business before initiating an action under this title.

Judge Anthony D. Johnstone:

Thank you.

Judge Mark W. Bennett:

I have one.

Judge M. Smith, Jr.:

Okay, go ahead.

Judge Mark W. Bennett:

Because of the Supreme Court's ruling, this case has a number of moving parts. If we were to hypothetically view the DPIA provision or provisions as unconstitutional, does the attorney general believe there would be any benefit to certifying to the California Supreme Court the severability question?

Kristin Liska:

That is certainly an option on the table if this court has serious concerns about severability. Again, we think the arguments that we make have made clear that the DPIA requirement would be severable, but if this court has deep concerns, that is always an option on the table.

Judge M. Smith, Jr.:

Very well. Okay. We'll give you a little extra time because we're taking over.

Kristin Liska:

Thank you, Your Honor.

Judge M. Smith, Jr.:

All right, so is it Corn-Revere, is that the correct pronunciation? Please proceed, sir.

Robert Corn-Revere:

Good morning, Your Honor, and may it please the Court. AB-2273 is a speech regulation masquerading as a privacy law. The overall thrust of the act is to regulate internet speech, how it is presented, its content, and who is connected.

Judge M. Smith, Jr.:

Counsel, we don't have a whole lot of time. Before we get too far along in this, as you well know, you have challenged this law facially. How does the Supreme Court's recent decision in Moody impact your challenge here, if at all?

Robert Corn-Revere:

I don't think it impacts our challenge at all. And as a matter of fact, I think it confirms that facial challenges in the First Amendment context are allowable when a substantial number of the applications of the law are unconstitutional compared to its plainly legitimate sweep. And I think that standard is clearly met here.

Judge M. Smith, Jr.:

Let's say, arguendo again–

Robert Corn-Revere:

Certainly.

Judge M. Smith, Jr.:

If I were to agree with you about the DPIA portion of the law, but not necessarily with respect to the rest of it, what would that require us to do? Can we basically just attack or deal with that one portion of it and not deal with the other if we felt the district court was incorrect in its facial analysis of the balance of the law?

Robert Corn-Revere:

Well, I think as Your Honor's questions indicated during my friend's presentation that the law would be very hard to understand or apply if the DPIA provision were knocked out. It ties directly to Section 35, which is the enforcement provision, which starts with the predicate in Section 35A that if an entity that is regulated by the act is in substantial compliance with the law, then you get a 90-day grace period. That presupposes that we know what substantial compliance means and it requires that substantial compliance in the first place.

Judge Mark W. Bennett:

Counsel, let me address one particular part of the statute, although this is not the limit of my concerns. I don't know if describing it as a subsection is correct, but you'll know what I mean. Section or Subsection 32, establishing a working group, the district court enjoined it, which to me is essentially a prior restraint on speech. The district court enjoined the speech that is called for in the statute, that is individuals making a report to the governor. So how is it that that, for example, setting up a working group wouldn't be severable even if the DPIA provisions were declared unconstitutional, even if, as the district court said, "Well, the working group might be opining on things we've already said are unconstitutional." But why wouldn't, for example, that be severable?

Robert Corn-Revere:

I think the working group would be severable except in a facial challenge as the Supreme Court said in Patel versus Los Angeles that you look at in analyzing the overbreadth of a statute, the regulatory parts of the statute. We don't view the working group as regulatory as an advisory panel for legislation.

Judge Mark W. Bennett:

But the judge enjoined it. The district judge enjoined the working group.

Robert Corn-Revere:

Yes, but I don't think that that as a – The fact that there's an advisory group provides a plainly legitimate sweep that would save the overall statute. If you look at the working parts, the regulatory parts–

Judge Mark W. Bennett:

But wouldn't it save the working group part?

Robert Corn-Revere:

Well, not to the extent that the working group is providing advice on how the law is to be implemented and applied, which was the purpose.

Judge Mark W. Bennett:

So a prior restraint by the federal court barring individuals appointed by the legislature from opining on putting a statute into effect that one or four judges have declared is problematic, that that's not legitimate or severable or couldn't be enforced?

Robert Corn-Revere:

It could be enforced, but it's not enough to save the statute. I think if you look at the parts that regulate speech, you can look at the law in essentially four buckets. The first of which you extensively discussed during my friend's time. Those are the DPIA requirements. I think they are plainly a content-based regulation. They are a prior restraint, as we explained in our papers. We agree with the district court except to the point that the court decided that this was-

Judge M. Smith, Jr.:

Following up on my colleague's question though, and again, this is hypothetical, if our court, arguendo, were to find the DPIA is unconstitutional and is not severable, do you even care whether we get to the balance of the law on the basis that it is in fact not facially challengable? Does it matter to you?

Robert Corn-Revere:

Well, it does because the other provisions are equally unconstitutional as the prior restraint in the DPIA. Look at Section 31 A5, the age assurance requirement. That requires every website, whether or not they previously had any mechanism for judging the age of its users to do something in that regard or limit the way it can deliver speech.

Judge Anthony D. Johnstone:

Robert Corn-Revere, before you get too much further, I'm eager to hear about the other three buckets which might help us structure our future questions.

Robert Corn-Revere:

And that is the second bucket. The first one is the DPIA requirements. The second one is age assurance. The third is the policy enforcement provision of Section 31 A9, which essentially deputizes private parties and gives the state oversight ability to determine whether or not they are properly applying their content-based standards. And as the evidence in the district court revealed, various subjective standards, editorial standards. And then the final are the data use provisions in Section 31 B1 through 7. Those are the ones that restrict how data can be used. And as we noted in our papers, those also restrict speech as the Supreme Court held in Sorrell v. IMS Health.

Judge M. Smith, Jr.:

I want to get back to, again, following on what we've all been talking about here, why, arguendo, would it be a problem for you if the court were to say the DPIA is unconstitutional, it's not severable, and that because of that the rest of the act is not enforceable either because you can't enforce it? Well, how about that? What's your position on that?

Robert Corn-Revere:

Well, we take the win–

Judge M. Smith, Jr.:

Of course.

Robert Corn-Revere:

... but I think there are so many things about this law that are unconstitutional that it's important for the court.

Judge M. Smith, Jr.:

But if we, again, arguendo, if we conclude particularly in light of Moody, that the balance of the law is not wholly facially challengable, it would have to be done on an as applied basis and considered section by section that you would have to send back to the district court again. Do you agree?

Robert Corn-Revere:

No, Your Honor. First of all, because of the rule established in Patel that you look at the regulatory portions of the law. But beyond that, the question, and it's repeated from the overbreadth analysis in United States v. Hansen that you look at the extent to which the law has unconstitutional applications and then you determine whether or not that is too extensive in relationship to the law's plainly legitimate sweep. And so you're looking at how much speech is suppressed by the unconstitutional portions compared to other provisions and whether or not that is sufficient to save the law.

Judge Anthony D. Johnstone:

But, Robert Corn-Revere, I guess this is a very helpful discussion trying to distinguish between the severability on the one hand, the facial on the other, and the section-by-section analysis. I guess I read Moody, and correct me where I'm wrong here, and I think I would read Hansen the same way to be distinguishing between a facial challenge to a single provision. I think in Hansen they're looking at the applications and there's some contextual work, but they're not actually scrutinizing the other pieces that inform the reading of the statute. In Moody, they're also slicing and dicing for the purpose of the facial challenge. In other words, why isn't the facial challenge to this subsection and that has to be dealt with? And then there's a separate facial challenge to is this subsection unconstitutional without a plainly legitimate speech? Why wouldn't we do it or require the district court to do it section by section in that way and then we're left with a severability analysis?

Robert Corn-Revere:

Well, there are a couple of responses to that, Your Honor. One is the district court in essence did go section-by-section. The state had requested a briefing on the severability question. There was a separate briefing on that, and the district court then did go section-by-section through the act and determined for purposes of the severability analysis, whether or not those provisions were constitutional or not. Of the ten mandates and eight prohibitions in the law, the court determined that eight of the mandates and five prohibitions violated the First Amendment.

Judge Anthony D. Johnstone:

So there's no legitimate sweep to asking a company to estimate the age of its consumers for purposes of non-content-based safety regulations.

Robert Corn-Revere:

Well, again, all of these regulations are tied to the content-based determination of whether or not this is safe. As you pointed out in looking at the factors that the court or that the state requires companies to assess when they're determining in their DPAIs–

Judge Anthony D. Johnstone:

Well, I did that with respect to the DPIA. I don't think that... I guess how is the age estimation, for example, tied to the factors in the DPIA?

Robert Corn-Revere:

Well, again, the factors in the DPIA determine what you were looking for and why you imposed the age determination and the age determination is being made to determine whether or not it is acceptable to communicate with the user and how that communication can take place, for example, whether you would recommend content.

Judge M. Smith, Jr.:

That really leaves us in my judgment with the same problem the Supreme Court dealt with in Moody. Because as have been pointed out by our questions, there are sections of the non-DPIA portion of the statute that at least in my judgment, cannot be analyzed on a facial basis. It requires an as-applied analysis. The district court did not do that, right?

Robert Corn-Revere:

Well, the district court didn't follow the formula that Moody prescribed because Moody came after.

Judge M. Smith, Jr.:

Which is why the Supreme Court sent it back.

Robert Corn-Revere:

But the law hasn't changed and there's a reason why the court sent it back in Moody where it would not be appropriate here. Keep in mind, in Moody, the court was looking at an analysis from the circuit courts that involved two services for two different online companies, basically newsfeed and those kinds of things. Whereas it did not look at all of the conceivable applications such as whether or not Uber is covered.

Judge Mark W. Bennett:

Counsel, respectfully, I think the district court's decision is a little bit different, and I'm looking at ER38, page 37 of the court's decision. Given that multiple provisions of the CAADCA will be preliminarily enjoined, and the court's determination that these are not functionally severable from the presumably valid remainder of the statute, the court concludes that it cannot sever the likely invalid portions from the statute and sustain the remainder. So the only basis that I see for at least a significant part of the court's ruling is the severability analysis and not doing the type of analysis that the Supreme Court, I think, has commanded in Moody. I believe that's what my colleague is saying, but I don't want to put words obviously in Judge Smith's mouth.

Judge M. Smith, Jr.:

You've done an excellent job of it.

Robert Corn-Revere:

I would respond in this way, and that is obviously Moody came after the District Court's opinion, so the district court didn't have the benefit of framing its analysis in those terms. But Moody didn't substantially change what the law is. It basically affirmed the analysis in Hansen, that if there are enough unconstitutional provisions of a law that it can be challenged as overbroad and challenged on its face. And you compare those findings to the plainly legitimate sweep, the district court's analysis, because it did go provision by provision in analyzing the severability, gives this court sufficient information on which to fit the analysis into the mold prescribed by Moody and it is de novo review-

Robert Corn-Revere:

Just into the mold by Moody.

Judge M. Smith, Jr.:

But counsel–

Robert Corn-Revere:

And it is a de novo review.

Judge M. Smith, Jr.:

As we're all saying here, though, as you very well know, a facial challenge, there can be no circumstance in which any application of the statute is constitutional. And this was pointed out with respect to the working group, for example, that probably has no problem at all. So if there are portions of the law that clearly don't fall on a facial analysis, what does that do with what we deal with here? DPIA may be a separate issue, but with respect to the balance of the statute, I don't think the district court applied anything other than the facial analysis, and it seems to be the very kind of thing that the Supreme Court said. And Moody was not right, it didn't fit in the facial analysis category.

Robert Corn-Revere:

Well, I would distinguish the facial analysis category in a case like Salerno from a facial analysis in a First Amendment challenge as in Stevens and in Moody. And in that case, you're looking at not just whether or not any provision can survive, but whether or not the unconstitutional provisions as determined by the court outweigh those provisions that could still exist.

Judge Mark W. Bennett:

So following up on what I think was one of Judge Johnston's questions. In doing that analysis, where one is putting the good on one side and the not good on the other and seeing if the not good substantially outweighs the good, that your view is we don't look at individual distinct provisions of the statute that because the California legislature has chosen to put 10 or 15 or 20 different provisions that deal with some different things with the same statute or bill number, that we look at all of them together and divide them up as opposed to looking at distinct statutory provisions in the bill, simply because the legislature chose to stick them all in one legislative enactment.

Robert Corn-Revere:

Well, no, I think you look at the overall thrust of what the bill regulates and then determine whether or not the unconstitutional portions outweigh that which is constitutional.

Judge Mark W. Bennett:

But you don't think in doing that we should analyze what we consider to be separate type provisions separately and look at Provision 29 and say we need to do a separate facial challenge analysis of 29? Because, yeah, it's got some unconstitutional stuff but there's plenty of section 29 that there are circumstances where that works and so a facial challenge doesn't work. You're saying we have to do one through 29 together, we can't look at seven separately from 29 in making our numerical sorting?

Robert Corn-Revere:

Well, no, I mean, you can go either way. You can look at the statute as a whole and determine how much speech it regulates, and in this case the answer is a vast amount of speech.

Judge Anthony D. Johnstone:

But why doesn't that confuse it with a severability analysis? That sounds like a question for severability, not a question for a facial challenge as to each of the law's provisions.

Robert Corn-Revere:

Well, I mean, I think they are interrelated in that if you're doing a severability analysis, you are making a judgment about each of the law's provisions and determining its constitutionality. At the end of that, if you determine that the overall impact of the law is to regulate a substantial amount of protected speech compared to its plainly legitimate sweep, then at that point you make the judgment on a facial basis that is unconstitutional.

Judge Anthony D. Johnstone:

So just to take another example, you don't challenge at least some of the geolocation provisions?

Robert Corn-Revere:

That's right, there were four provisions in section 31B we did not challenge.

Judge Anthony D. Johnstone:

Okay. This matters because the first amendment is federal law, the severability question is state law, so I think it's pretty important that we keep those two separate within our jurisdiction.

Robert Corn-Revere:

Yeah.

Judge Anthony D. Johnstone:

That's not even challenged. So you're saying that, simply by virtue, as Judge Bennett says, of the fact that the legislature has decided to compound these subject to under state law its own severability analysis, the unchallenged geolocation of children protections in the bill fail simply because it was enacted alongside the more suspect DPIA.

Robert Corn-Revere:

Well, and therein lies the problem, because–

Judge Anthony D. Johnstone:

As a matter of First Amendment law, not severability law.

Robert Corn-Revere:

Right. If having any unchallenged or any constitutional provisions are enough to immunize a law from overbreadth challenge, then you've simply made overbreadth an anachronistic doctrine that it simply can't be applied.

Judge M. Smith, Jr.:

I gather what you're saying is that, even if the district court failed... Well, in the way he had applied a facial challenge to the balance of the law, not the DPIA portion, it's okay to consider it for purposes of severability, but if not for purposes of First Amendment analysis. Is that correct?

Robert Corn-Revere:

I'm not sure I understand your question.

Judge M. Smith, Jr.:

In other words, the district court... Let's go back. If the DPIA is, in quotes, unconstitutional, in order for us to determine if we did that it is not severable, can we look at the balance of the act? Even if we think the district court could not analyze it facially, can we look at the balance of it and say, under California law, this isn't inseverable? Can we do that?

Robert Corn-Revere:

Yes, absolutely. I mean, you can-

Judge M. Smith, Jr.:

Is that what the district court in effect did?

Robert Corn-Revere:

Yeah. And the district court did determine that that portion was not severable because it was interrelated with the rest of the way the law operated. Let me just add this, and that is, if the court is inclined to have doubts about whether or not this is subject to a facial over breath challenge, I would request the opportunity for additional briefings since Moody did come after the district court's decision. I would ask also that, if there is any further consideration or even a remand, that the injunction remain in place, because what the state has never contested or the factors involved in granting preliminary injunctive relief and that is irreparable harm and likelihood of success. Those–

Judge M. Smith, Jr.:

Your time is up. If we want additional briefing, we will certainly let you know. But thank you very much for your argument. So the state has some rebuttal time and we're going to give you an extra minute because we took you over, and I know you want that minute, right?

Kristin Liska:

Of course. Thank you, Your Honor. Appreciate that. I think I'd like to pick up on where the discussion with my friend on the other side was going, which is that, when you look at this law into a facial challenge, we would contend you need to do a facial analysis of each separate subsection, and that was really what the lower court failed to do. The question, and Moody it makes clear, is not what is the overall thrust of the law. I mean, if you look at the language in Moody, it says a court needs to look at, "what activities by what actors do the laws prohibit or otherwise regulate," and drill down into this full set of applications.

So we think that really in a law this complex it necessitates looking at each separate provision because... And what the law court did is it just grouped them together and said, "Speech is involved." But if you look and disentangle these, that's not necessarily the case. And different constitutional tests may apply to different provisions. Compelled speech, for instance. We argue that if you look at the provision about enforcing terms and services, that's not a First Amendment problem under cases like Cohen that relate to private agreements. Parts like age estimation–

Judge M. Smith, Jr.:

Let me ask you what I asked your colleague because it's important to me. The district court in its severability analysis, again a state law question, looked at the entirety of the bill. For purposes of analyzing what we do with this injunction, can we argue Endo, if we said the DPIA is unconstitutional, even if we disagree with the way the district court analyzed the balance of the bill for purposes of First Amendment law, can we interpret in determining whether this DPIA section is severable? Can we analyze it based on the district court's review of the impact of the balance of the legislation?

Kristin Liska:

I think that, yes, I want to say that I track the way that you're viewing this. I think the way to view severability is if you were to take an entire piece of paper and you were to carve out a hole from the middle, which is the part that you view as unconstitutional, can the rest of the piece of paper stand? So you do have to look at the remainder of the law with respect to, if I pull out that part that is unconstitutional, does the remainder collapse like a house of cards or is it sturdy enough to stand? I guess maybe a Jenga tower is a more apt metaphor. Is it like a Jenga tower where I pulled out a block and the rest of the tower stands very well? And that's the three prong analysis under California law. Is it grammatically separable, operationally separable in that the rest of the law can function without the piece?

And is it volitionally separable? Which we note in our reply brief asks, would the legislature have preferred the remainder of the law to nothing? Not would it have preferred everything it could have gotten versus the part taken out, but would it have preferred to keep everything it could keep constitutionally to having nothing in place? So we do think you have to look at the entire law to do the severability analysis, but it starts from what is the specific provision that's unconstitutional and we think the district court's error was not drilling down at each subdivision to determine that, especially under Moody. I see my time's up. Thank you, Your Honor.

Judge M. Smith, Jr.:

Any questions? I'm sure we could talk about this for a really, really long time, but we're not going to. But thanks for your learned argument. We appreciate it very much. The case just argued is submitted.