The bill is already bad enough, but the op-ed in support of it is even more ridiculous. It pulls on a bunch of typical techno moral panic tropes, none of which is supported by the evidence. Let’s start at the beginning:
Kids are even more in the bag of social media companies than we think. So many of them have ceded their online autonomy so fully to their phones that they even balk at the idea of searching the internet — for them, the only acceptable online environment is one customized by big tech algorithms, which feed them customized content.
This is the classic techno moral panic trope that the technology is some sort of brain eating mind control system, creating zombies who have no free will of their own.
Is there any actual evidence to support this? Nope. This mind control trope is quite popular in moral panics of the past, though. We’ve seen it in panics about television and movies. The “satanic panic” was basically this on steroids. Panics about things like “dungeons and dragons” fit the same bill as well. Common among this trope is the “warning” from some wise politician who seems positive that everyone else (especially children) are somehow powerless against the power of the new tech to take over their minds.
It’s an inherently obnoxious and elitist position, basically claiming “the populace it too fucking stupid, and too easily led to think for themselves,” so clearly they’ll outsource their thinking to the dreaded algorithms.
That there’s literally zero evidence supporting this doesn’t matter to someone like Senator Murphy. It sounds plausible, and so the narrative wins the day.
The truth, of course, is that as social media has become more popular the actual data suggests it has mostly helped kids, not hurt them. But facts like that are too inconvenient for the narrative.
And, what you lack for in statistical evidence, you can always replace with random anecdote, as Murphy does. He uses some kids he met with at a high school as a prop to suggest… well… you have to read it for yourself:
This spring, I visited with a group of high school students in suburban Connecticut to have a conversation about the role that social media plays in their daily lives and in their mental health. More children today report feeling depressed, lonely and disconnected than ever before. More teens, especially teen girls and L.G.B.T.Q. teens, are seriously considering suicide. I wanted to speak candidly about how social media helps and hurts mental health. By the end of the 90-minute dialogue, I was more worried than ever about the well-being of our kids — and of the society they will inherit.
There are numerous problems with children and adolescents using social media, from mental health deterioration to dangerous and age-inappropriate content and the lackluster efforts tech companies employ to enforce their own age verification rules. But the high schoolers with whom I met alerted me to an even more insidious result of minors’ growing addiction to social media: the death of exploration, trial and error and discovery. Algorithmic recommendations now do the work of discovering and pursuing interests, finding community and learning about the world. Kids today are, simply put, not learning how to be curious, critical adults — and they don’t seem to know what they’ve lost.
This again picks up on a mind control trope. I remember fears when I was in high school that letting kids use calculators would stop them from being able to truly understand math. The idea that social media leads to “the death of exploration, trial and error and discovery” is laughable.
The internet is the greatest tool for exploration, trial and error, and discovery ever invented. Teens discover all sorts of things by going through the internet. And, to the extent that algorithms play any role at all, it’s often to help them find more of what it is they want to explore. I know kids who have “explored” and “discovered” all sorts of things from learning about history, to how to solve Rubik’s cubes, to how to improve swimming techniques, to basically every topic under the sun.
It is simply bizarre and wholly unsupported. Indeed, studies have suggested that internet access for kids has tends to make kids smarter and more knowledgeable. Yes, there are risks, but the research keeps suggesting that the way to deal with that is not to ban algorithms or ban kids from social media, but to involve parents and educators to teach kids how to properly and safely use the internet, and what to do when they come across problematic content (rather than trying to hide the existence of problematic content from them entirely).
Murphy, however, paints a picture of the internet that basically does not exist:
Their dependence on technology sounds familiar to most of us. So many of us can barely remember when we didn’t have Amazon to fall back on when we needed a last-minute gift or when we waited by the radio for our favorite songs to play. Today, information, entertainment and connection are delivered to us on a conveyor belt, with less effort and exploration required of us than ever before.
What? I mean, if I’m using Amazon for a last minute gift, I still have to search and type in a bunch of stuff to figure out what it is I want to buy. I don’t just purchase the first thing that comes up on Amazon’s front page. And, as for music, I actually spend a fair bit of time searching through playlists and whatnot to find interesting music.
Besides, Murphy’s pining for history when… you had to “wait for our favorite song to play” on the radio… is, um, even worse than what he describes today. If his concern is that with today’s algorithms, you have no influence, and, say, make no effort in “discovery,” then having to rely on DJs on one of the three stations you can hear from your house is way worse. Today, we can explore, and experiment, and listen to different genres and styles, and if we hear something we like, we can ask for more like that, or look at other similar artists. In other words, in today’s world, thanks to algorithms, we actually have an even greater ability to discover and to explore and to engage in trial and error.
The world Murphy describes is literally the opposite of today. And yet, he immediately follows up this pining for the world in which you had to wait for the DJ to play a song you liked with:
A retreat from the rituals of discovery comes with a cost. We all know instinctively that the journeys in life matter just as much as the destinations. It’s in the wandering that we learn what we like and what we don’t like. The sweat to get the outcome makes the outcome more fulfilling and satisfying.
Dude, what? Again, when you had to wait for the radio to play songs you liked, you got to engage in no real discovery. The “algorithm” then was what the DJ wanted to play (or, if in more recent years, what the giant radio conglomerate wanted to play). With modern music systems on the internet, we can actually engage in real discovery. It’s bizarre and clearly nonsense that Murphy thinks the old way involved more discovery.
Why should students put in the effort to find a song or a poem they like when an algorithm will do it for them?
Again, people always search out what they like. The internet makes it possible for them to do so. I could just as easily point out that that in the past, kids wouldn’t be able to “find a song” they liked because they were reliant on a DJ choosing what would play on the radio. The world is much more open to discovery and experimentation today.
Why take the risk to explore something new when their phones will just send them never-ending content related to the things that already interest them?
Um, mainly because lots of people like to explore new things, and it’s much easier and lower risk to do so today than in the past. Murphy and I are similar in age, and my childhood exploration of finding music sucked. I would spend hours in record stores and have to spend $15 to $20 on random albums that I had no idea if I would actually like. It was not a fun process. It sucked. I wasted a ton of money, and didn’t feel like I was “discovering” much. Today is so amazing for music discovery.
Sure, music services recommend stuff to me, but I spend a ton of time exploring playlists, and different bands, and can try out all sorts of things. None of that was possible when I was a kid.
There’s a lot more in the article, but it’s just more of the same moral panic nonsense. Murphy has become the old man yelling at the clouds about the kids today and their calculators. It’s not based on reality. It’s based on fear and nonsense.
There are some questions about whether or not Section 230 protects AI companies from being liable for the output from their generative AI tools. Matt Perrault published a thought-provoking piece arguing that 230 probably does not protect generative AI companies. Jess Miers, writing here at Techdirt, argued the opposite point of view (which I found convincing). Somewhat surprisingly, Senator Ron Wyden and former Rep. Chris Cox, the authors of 230 have agreed with Perrault’s argument.
The Wyden/Cox (Perrault) argument is summed up in this quote from Cox:
“To be entitled to immunity, a provider of an interactive computer service must not have contributed to the creation or development of the content at issue,” he told me. “So when ChatGPT creates content that is later challenged as illegal, Section 230 will not be a defense.”
At a first pass, that may sound compelling. But, as Miers noted in her piece, the details get a lot trickier once you start looking at them. As she points out, it’s already well established that 230 protects algorithmic curation and promotion (this was sorta, partly, at issue in the Gonzalez case, though by the time the Supreme Court heard the case, it was mostly dropped, in part because the lawyers backing Gonzalez realized that their initial argument probably would make search engines illegal).
Further, Miers notes, that 230 cases have already been found to protect algorithmically generated snippets that summarize content elsewhere, even though those are “created” by Google, based on (1) the search input “prompt” from the user, and (2) the giant database of content that Google has scanned.
And, that’s where the issue really gets tricky, and where those insisting that generative AI companies are clearly outside the scope of 230 feel like they haven’t quite thought through all of this: where is the line that you can draw between these two things? At what point do we go from one tool, Google, that scraped a bunch of content and creates a summary in response to input, to another tool, AI, that scrapes a bunch of content and creates “whatever” in response to input?
Well, the two Senators who hate the internet more than anyone else, the bipartisan “destroy the internet, and who cares what damage it does” buddies: Senator Richard Blumenthal and insurrectionist supporting Senator Josh Hawley have teamed up to introduce a bill that explicitly says AI companies get no 230 protection. Leaving aside the question of why any Democrat would be willing to team up with Hawley on literally anything at this moment, this bill is… well… weird.
First, just the fact that they had to write this bill suggests (perhaps surprisingly?) that Hawley and Blumenthal agree with Miers more than they agree with Wyden, Cox, or Perrault. If 230 didn’t apply to AI companies, why would they need to write this bill?
But, if you look at the text of the bill, you quickly realize that Hawley and Blumenthal (this part is not surprising) have no clue how to draft a bill that wouldn’t suck in a ton of other services, and strip them of 230 protections (perhaps that’s their real goal, as both have tried to destroy Section 230 going back many years).
The definition of “Generative Artificial Intelligence” is, well, a problem:
GENERATIVE ARTIFICIAL INTELLIGENCE.—The term ‘generative artificial intelligence’ means an artificial intelligence system that is capable of generating novel text, video, images, audio, and other media based on prompts or other forms of data provided by a person.’
First off, AI is quickly getting built into basically everything these days, so this definition is going to capture much of the internet within a few years. But, going back to the search example discussed above. Where the courts had said that 230 protected Google’s algorithmically generated summaries.
With this bill in place, that’s likely no longer true.
Or… as social media tools build in AI (which is absolutely coming) to help you craft better content, do all of those services then lose 230 protection? Just for helping users create better content?
And, of course, all of this confuses the point of Section 230, which, as we keep explaining, is just a procedural fast pass to get frivolous cases tossed out.
Just to make this point clear, let’s look at what happens should this bill become law. Say someone does a Google search on something, and finds that the automatically generated summary is written in a way that they feel is defamatory, even though it’s just a computerized attempt to summarize what others have written, in response to a prompt. The person sues Google, which is no longer protected by 230.
With Section 230, Google would be able to get the case kicked out with minimal hassle, as they’d file a relatively straightforward motion to dismiss pointing to 230 and get the case dismissed. Without that, they can still argue that the case is bad because, as an algorithm, Google could not have had the requisite knowledge to say anything defamatory. But, this is a more complicated (and more expensive) legal argument to make, and one that might not get tossed out on a motion to dismiss, but which would have to go through discovery, and to the more involved summary judgment stage, if not go all the way to trial.
In the end, it’s likely that Google still wins the case, because it had no knowledge at all as to whether the content was false, but now the process is expensive and wasteful. And, maybe it doesn’t matter for Google, which has buildings full of lawyers.
But, it does matter for basically every AI startup out there. Or any other company making use of AI to make their products better and more useful. If those products spew out some nonsense, even if no one believes it, must we have to fight a court battle over it?
Think back to the case we just recently spoke about regarding OpenAI being sued for defamation. Yes, ChatGPT appeared to make up some nonsense, but there remains no indication that anyone believed the nonsense. Only the one reporter saw it, and seemed to recognize it was fake. If he had then published the content, perhaps he would be liable for spreading something he knew was fake. But if it’s just ChatGPT writing it in response to that guy’s prompts, where is the harm?
In other words, even in the world of generative AI, there are still humans in the loop, and thus there can still be liability placed on the party responsible for (1) creating, via their prompts, and (2) spreading (if they publish it more widely) the violative content.
It still makes sense, then, for 230 to protect the AI tools.
Without that, what would AI developers do? How do you train an AI tool to never get anything wrong in producing content? And, even if you had some way to do that, wouldn’t that ruin many uses of AI? Lots of people use AI to deliberately generate fiction. I keep hearing about writers using it as a brainstorming tool. But if 230 doesn’t protect AI, then it would be way too risky for any AI tool to even offer to create “fiction.”
Yes, generative AI feels new and scary. But again, this all feels like an overreaction. The legal system today, including Section 230, seems pretty well equipped to handle specific scenarios that people seem most concerned about.
The Supreme Court is currently deliberating whether or not algorithms deserve protections under Section 230. And I hear from lots of people that maybe Section 230 wasn’t meant to cover algorithmic policing and recommendations of content. But that’s utter nonsense.
The whole area of content moderation first came about as a response to the earliest versions of spam. And one thing that people learned quite quickly is that you can’t manually police for spam if your site has even the slightest level of popularity. It will get flooded.
This is why any site needs to have some sort of automation to deal with spam. Indeed, for years, Techdirt has actually been using a combination of multiple different tools and setups to fight spam comments, but I’d never really looked into the numbers until just recently (mostly on a whim because I found the setting where those stats are!), and it’s kinda stunning. First off, we get way more spam attempts than I had even realized.
In the last six months alone, Techdirt received over 1.3 million attempts to spam our comments. That’s compared to the slightly over 40,000 legitimate comments we received in the same period. Here’s a chart of the spam comments per month:
I have no idea why spam grew so rapidly in January and February before falling in March, but even with 125,000+ spam messages in March, it completely dwarfs the amount of legitimate comments we got. The only possible way to keep up is to use automated systems. And, to be clear, while some small percentage of spam does get through, and we have a few legit comments caught in the spam filter, I’d argue we do a pretty good job of catching most spam, and allowing through most comments.
But, the larger point: without the multiple algorithmic systems we use to catch spam, we’d never be able to manage that amount of spam. Hell, we couldn’t handle manually dealing with less than 1% of the spam we actually get attempted right now. It would overwhelm us.
So if the courts (or, horror of horrors, Congress) were to decide that “algorithms” no longer are protected under Section 230, it would destroy Techdirt. While the 1st Amendment would eventually protect us, the lack of 230 protections would make using a spam filter a liability that would open up the risk of having to fight a full legal battle just to prove our right to block spam comments.
As such, our choices would be to turn off the algorithms and let spam flow, shut down our comments entirely, or risk ruinous lawsuits for the “harm” of trying to stop spam with an automated filter.
Technological filters (i.e., algorithms) should obviously be protected by Section 230, because without them, we lose the ability to fight spam, and the amount of such content is truly overwhelming. And that’s just for us, a pretty small site. Imagine how larger sites are dealing with this stuff.
But now it turns out that, all along, he’s set up a special “shadow boosting” system, that allows him to pump his own personal favorite accounts into your algorithmically generated “for you” feed all the time.
Now, we already know that Elon’s own tweets got the “max boost” treatment from engineers after Elon had a sad over a Joe Biden tweet getting more engagement. But we were told that was a special treatment just to keep the guy in charge happy.
However, it probably won’t surprise most of you to know that there’s also a special list of “VIPs” whose tweets are boosted to appear in people’s feeds basically all the time. And the folks at Platformer got their hands on the list.
But Twitter does have a different standard for celebrities – including Musk himself. For months, the platform has maintained a list of around 35 VIP users whose accounts it monitors and offers increased visibility alongside Elon Musk, according to documents obtained by Platformer. The list, which spans the political gamut and also includes several journalists and celebrities, includes:
NBA All-Star LeBron James
Daily Wire founder and conservative commentator Ben Shapiro
Pseudonymous conservative commentator @catturd2
Rep. Alexandria Ocasio-Cortez, D-NY
President Joe Biden
YouTube star MrBeast
Venture capitalist and Twitter investor Marc Andreessen
Weird Twitter pioneer @dril
Comedian Jaboukie Young-White
Tesla community account @teslaownerssv
Journalists Matt Yglesias, Glenn Greenwald, Noah Smith, and Adrian Wojnarowski
(Platformer is not publishing the full list, whose makeup has changed slightly over time, to protect our sources’ identities. All the names above are still on the list.)
That sure seems to go against his “no lords and peasants” dual class system, or his Shatner spat about treating everyone equally, but since when has Elon ever been honest about anything?
I’ve seen some people surprised at who is on this list, but it primarily seems like a list of people that Elon thinks drive engagement.
In the meantime, there are some other changes happening at Twitter as well, including firing more of the remaining trust & safety team, and handing over more power to the very same AI that Musk seems to spend every day on Twitter mocking as dangerous. I’m sure that’ll work out great:
In an effort to save money, Twitter is scaling back its content moderation team even further, and relying more heavily on automated systems to police content than ever before.
Still, perhaps even bigger is Musk’s new plan, which is hilarious in just how much it demonstrates how little Musk understands about how Twitter works for everyone but himself. He announced that only people who pay (which he falsely calls “verified”) will show in the “for you” algorithmic feed that Elon’s Twitter now forces on every user when they open up Twitter, whether they like it or not:
I mean, almost all of this is hilarious. First, it makes the “for you” feed even more ridiculous and less valuable to users, meaning that it actually devalues the benefits of paying for Twitter Blue. Basically, it now means that appearing in the “for you” feed means you’re a chump.
But, once again, he’s creating a “lords and peasants” scenario, where the “lords” are simply those people gullible enough to pay Elon Musk a monthly fee.
Still, more importantly, the idea that he thinks this is the “only realistic way” of dealing with “AI bot swarms” really says a lot, and none of it good. First off, Elon had already claimed that he (1) took over Twitter to get rid of the bots and had such a good plan to do so that (2) he claimed the problem was already taken care of back in December, which he accomplished by… shutting off access to big telecom providers in India, Russia and Indonesia, and accidentally blocking a bunch of legit Twitter accounts.
Guess not.
Also, for scammers who are willing to pay $8, they’ll now have a clearer field to do their scamming, which could easily be worth more than $8 to the scammers.
And, yeah, the voting in polls thing: I mean, what? Most people are assuming that he’s still bitter about “losing” the poll about whether or not he should step down as CEO, which he promised to abide by even as it’s unclear that he’s taken any steps towards that end result. Even if it were true that “AI bot swarms” were polluting polls… so what? These polls mostly don’t matter.
To be fair, he also seems to be claiming that these new bots are “AI bot swarms” that have figured out how to bypass CAPTCHAs.
So, this seems to be his — I guess some might call it “strategy?” — for dealing with this. This is just one step short of saying that you can’t tweet unless you pay, which… will just drive a huge segment of the Twitter userbase away from Twitter.
If there’s one consistent thing we’ve seen with Elon is that he seems simply incapable of considering literally anyone else’s experience on his site beyond himself. Every single move has been about improving his personal experience on Twitter, usually at the expense of everyone else’s.
He bans accounts that make him feel uncomfortable.
He tweaks the algorithm to promote his own content.
He changes the rules to protect his friends and friends of friends while banning people his friends dislike.
He attacks and ridicules those who challenge him.
It’s just consistently about his own world. So now he’s promoting tweets of people he likes, while making sure that the “algorithm” that he sees every time he logs in is only filled with fans so obsessed with Musk they’re willing to pay him a monthly fee to tweet in his general direction.
That’s… one way to run a social media network. But it doesn’t seem like a very good one.
Question Presented: Does Section 230 Protect Generative AI Products Like ChatGPT?
As the buzz around Section 230 and its application to algorithms intensifies in anticipation of the Supreme Court’s response, ‘generative AI’ has soared in popularity among users and developers, begging the question: does Section 230 protect generative AI products like ChatGPT? Matt Perault, a prominent technology policy scholar and expert, thinks not, as he discussed in his recently published Lawfare article: Section 230 Won’t Protect ChatGPT.
Perault’s main argument follows as such: because of the nature of generative AI, ChatGPT operates as a co-creator (or material contributor) of its outputs and therefore could be considered the ‘information content provider’ of problematic results, ineligible for Section 230 protection. The co-authors of Section 230, former Representative Chris Cox and Sen. Ron Wyden, have also suggested that their law doesn’t grant immunity to generative AI.
I respectfully disagree with both the co-authors of Section 230 and Perault, and offer the counter argument: Section 230 does (and should) protect products like ChatGPT.
It is my opinion that generative AI does not demand exceptional treatment. Especially since, as it currently stands, generative AI is not exceptional technology; an understandably provocative take to which we’ll soon return.
But first, a refresher on Section 230.
Section 230 Protects Algorithmic Curation and Augmentation of Third-Party Content
Recall that Section 230 says websites and users are not liable for the content they did not create, in whole or in part. To evaluate whether the immunity applies, the Barnes v. Yahoo! Court provided a widely accepted three-part test:
The defendant is an interactive computer service;
The plaintiff’s claim treats the defendant as a publisher or speaker; and
The plaintiff’s claim derives from content the defendant did not create.
The first prong is not typically contested. Indeed, the latter prongs are usually the flashpoint(s) of most Section 230 cases. And in the case of ChatGPT, the third prong seems especially controversial.
Section 230’s statutory language states that a website becomes an information content provider when it is “responsible, in whole or in part, for the creation or development” of the content at issue. In their recent Supreme Court case challenging Section 230’s boundaries, the Gonzalez Petitioners assert that the use of algorithms to manipulate and display third-party content precludes Section 230 protection because the algorithms, as developed by the defendant website, convert the defendant into an information content provider. But existing precedent suggests otherwise.
For example, the Court in Fair Housing Council of San Fernando Valley v. Roommate.com (aka ‘the Roommates case’)—a case often invoked to evade Section 230—held that it is not enough for a website to merely augment the content at issue to be considered a co-creator or developer. Rather, the website must have materially contributed to the content’s alleged unlawfulness. Or, as the majority put it, “[i]f you don’t encourage illegal content, or design your website to require users to input illegal content, you will be immune.”
The majority also expressly distinguished Roomates.com from “ordinary search engines,” noting that unlike Roommates.com, search engines like Google do not use unlawful criteria to limit the scope of searches conducted (or results delivered), nor are they designed to achieve illegal ends. In other words, the majority suggests that websites retain immunity when they provide neutral tools to facilitate user expression.
While “neutrality” brings about its own slew of legal ambiguities, the Roommates Court offers some clarity suggesting that websites with a more hands-off approach to content facilitation are safer than websites that guide, encourage, coerce, or demand users produce unlawful content.
For example, while the Court rejected Roommate’s Section 230 defense for its allegedly discriminatory drop-down options, the Court simultaneously upheld Section 230’s application to the “additional comments” option offered to Roommates.com users. The “additional comments” were separately protected because Roommates did not solicit, encourage, or demand their users provide unlawful content via the web form. In other words, a blank web form that simply asks for user input is a neutral tool, eligible for Section 230 protection, regardless of how the user actually uses the tool.
The Barnes Court would later reiterate the neutral tools argument, noting that the provision of neutral tools to carry out what may be unlawful or illicit content does not amount to ‘development’ for the purposes of Section 230. Hence, while the ‘material contribution’ test is rather nebulous (especially for emerging technologies), it is relatively clear that a website must do something more than just augmenting, curating, and displaying content (algorithmically or otherwise) to transform into the creator or developer of third-party content.
The Court in Kimzey v. Yelp offers further clarification:
“the material contribution test makes a “‘crucial distinction between, on the one hand, taking actions (traditional to publishers) that are necessary to the display of unwelcome and actionable content and, on the other hand, responsibility for what makes the displayed content illegal or actionable.’”).”
So, what does this mean for ChatGPT?
The Case For Extending Section 230 Protection to ChatGPT
In his line of questioning during the Gonzalez oral arguments, Justice Gorsuch called into question Section 230’s application to generative AI technologies. But before we can even address the question, we need to spend some time understanding the technology.
Products like ChatGPT use large language models (LLMs) to produce a reasonable continuation of human-sounding responses. In other words, as discussed here by Stephen Wolfram, renown computer scientist, mathematician, and creator of WolframAlpha, ChatGPT’s core function is to “continue text in a reasonable way, based on what it’s seen from the training it’s had (which consists in looking at billions of pages of text from the web, etc).”
While ChatGPT is impressive, the science behind it is not necessarily remarkable. Computing technology reduces complex mathematical computations into step-by-step functions that the computer can then solve at tremendous speeds. As humans, we do this all the time, just much slower than a computer. For example, when we’re asked to do non-trivial calculations in our heads, we start by breaking up the computation into smaller functions on which mental math is easily performed until we arrive at the answer.
Tasks that we assume are fundamentally impossible for computers to solve are said to involve ‘irreducible computations’ (i.e. computations that cannot be simply broken up into smaller mathematical functions, unaided by human input). Artificial intelligence relies on neural networks to learn and then ‘solve’ said computations. ChatGPT approaches human queries the same way. Except, as Wolfram notes, it turns out that said queries are not as sophisticated to compute as we may have thought:
“In the past there were plenty of tasks—including writing essays—that we’ve assumed were somehow “fundamentally too hard” for computers. And now that we see them done by the likes of ChatGPT we tend to suddenly think that computers must have become vastly more powerful—in particular surpassing things they were already basically able to do (like progressively computing the behavior of computational systems like cellular automata).
But this isn’t the right conclusion to draw. Computationally irreducible processes are still computationally irreducible, and are still fundamentally hard for computers—even if computers can readily compute their individual steps. And instead what we should conclude is that tasks—like writing essays—that we humans could do, but we didn’t think computers could do, are actually in some sense computationally easier than we thought.
In other words, the reason a neural net can be successful in writing an essay is because writing an essay turns out to be a “computationally shallower” problem than we thought. And in a sense this takes us closer to “having a theory” of how we humans manage to do things like writing essays, or in general deal with language.”
In fact, ChatGPT is even less sophisticated when it comes to its training. As Wolfram asserts:
“ChatGPT as it currently is, the situation is actually much more extreme, because the neural net used to generate each token of output is a pure “feed-forward” network, without loops, and therefore has no ability to do any kind of computation with nontrivial “control Flow.””
Put simply, ChatGPT uses predictive algorithms and an array of data made up entirely of publicly available information online to respond to user-created inputs. The technology is not sophisticated enough to operate outside of human-aided guidance and control. Which means that ChatGPT (and similarly situated generative AI products) are functionally akin to “ordinary search engines” and predictive technology like autocomplete.
Now we apply Section 230.
For the most part, the courts have consistently applied Section 230 to algorithmically generated outputs. For example, the Sixth Circuit in O’Kroley v. Fastcase Inc. upheld Section 230 for Google’s automatically generated snippets that summarize and accompany each Google result. The Court notes that even though Google’s snippets could be considered a separate creation of content, the snippets derive entirely from third-party information found at each result. Indeed, the Court concludes that contextualization of third-party content is in fact a function of an ordinary search engine.
Similarly, in Obado v. Magedson, Section 230 applies to search result snippets. The Court says:
Plaintiff also argues that Defendants displayed through search results certain “defamatory search terms” like “Dennis Obado and criminal” or posted allegedly defamatory images with Plaintiff’s name. As Plaintiff himself has alleged, these images at issue originate from third-party websites on the Internet which are captured by an algorithm used by the search engine, which uses neutral and objective criteria. Significantly, this means that the images and links displayed in the search results simply point to content generated by third parties. Thus, Plaintiff’s allegations that certain search terms or images appear in response to a user-generated search for “Dennis Obado” into a search engine fails to establish any sort of liability for Defendants. These results are simply derived from third-party websites, based on information provided by an “information content provider.” The linking, displaying, or posting of this material by Defendants falls within CDA immunity.
The Court also nods to Roommates:
“None of the relevant Defendants used any sort of unlawful criteria to limit the scope of searches conducted on them; “[t]herefore, such search engines play no part in the ‘development’ of the unlawful searches” and are acting purely as an interactive computer service…
The Court goes further, extending Section 230 to autocomplete (i.e. when the service at issue uses predictive algorithms to suggest and preempt a user’s query):
“suggested search terms auto-generated by a search engine do not remove that search engine from the CDA’s broad protection because such auto-generated terms “indicates only that other websites and users have connected plaintiff’s name” with certain terms.”
Like Google Search, ChatGPT is entirely driven by third-party input. In other words, ChatGPT does not invent, create, or develop outputs absent any prompting from an information content provider (i.e. a user). Further, nothing on the service expressly or impliedly encourages users to submit unlawful queries. In fact, OpenAI continues to implement guardrails that force ChatGPT to ignore requests that would demand problematic and / or unlawful responses. Compare this to Google Search which may actually still provide a problematic or even unlawful result. Perhaps ChatGPT actually improves the baseline for ordinary search functionality.
Indeed, ChatGPT essentially functions like the “additional comments” web form in Roommates. And while ChatGPT may “transform” user input into a result that responds to the user-driven query, that output is entirely composed of third-party information scraped from the web. Without more, this transformation is simply an algorithmic augmentation of third-party content (much like Google’s snippets). And as discussed, algorithmic compilations or augmentations of third-party content are not enough to transform the service into an information content provider (e.g. Roommates; Batzel v. Smith; Dyroff v. The Ultimate Software Group, Inc.; Force v. Facebook).
The Limit Does Exist
Of course, Section 230’s coverage is not without its limits. There’s no doubt that future generative AI defendants, like OpenAI, will face an uphill battle in persuading a court. Not only do defendants have the daunting challenge of explaining generative AI technologies for less technologically savvy judges, the current judicial swirl around Section 230 and algorithms does defendants no favors.
For example, the Supreme Court could very well hand-down a convoluted opinion in Gonzalez that introduces ambiguity as to when Section 230 applies to algorithmic curation / augmentation. Such an opinion would only serve to undermine the precedence discussed above. Indeed, future defendants may find themselves embroiled in convoluted debate about AI’s capacity for neutrality. In fact, it would be intellectually dishonest to ignore emerging common law developments that preclude Section 230 from claims alleging dangerous / defective product designs (e.g. Lemmon v. Snap, A.M. v. Omegle, Oberdorf v. Amazon).
Further, the Fourth Circuit’s recent decision in Henderson v. Public Data could also prove to be problematic for future AI defendants as it imposes contributive liability for publisher activities that go beyond those of “traditional editorial functions” (which could include any and all publisher functions done via algorithms).
Lastly, as we saw in the Meta / DOJ settlement regarding Meta’s discriminatory practices involving algorithmic targeting of housing advertisements, AI companies cannot easily avoid liability when they materially contribute to the unlawfulness of the result. If OpenAI were to hard-code ChatGPT with unlawful responses, Section 230 will likely be unavailable. However, as you might imagine, this is a non-trivial distinction.
Public Policy Demands Section 230 Protections for Generative AI Technologies
Section 230 was initially established with the recognition that the online world would undergo frequent advancements, and that the law must accommodate these changes to promote a thriving digital ecosystem.
Generative AI is the latest iteration of web technology that has enormous potential to bring about substantial benefits for society and transform the way we use the Internet. And it’s already doing good. Generative AI is currently used in the healthcare industry, for instance, to improve medical imaging and to speed up drug discovery and development.
As discussed, courts have developed precedence in favor of Section 230 immunity for online services that solicit or encourage users to create and provide content. Courts have also extended the immunity to online services that facilitate the submission of user-created content. From a legal standpoint, generative AI tools are not unique from any other online service that encourages user interaction and contextualizes third-party results.
From a public policy perspective, it is crucial that courts uphold Section 230 immunity for generative AI products. Otherwise, we risk foreclosing on the technology’s true potential. Today, there are tons of variations of ChatGPT-like products offered by independent developers and computer scientists who are likely unequipped to deal with an inundation of litigation that Section 230 typically preempts.
In fact, generative AI products are arguably more vulnerable to frivolous lawsuits because they depend entirely upon whatever query or instructions its users may provide, malicious or otherwise. Without Section 230, developers of generative AI services must anticipate and guard against every type of query that could cause harm.
Indeed, thanks to Section 230, companies like OpenAI are doing just that by providing guardrails that limit ChatGPT’s responses to malicious queries. But those guardrails are neither comprehensive nor perfect. And like with all other efforts to moderate awful online content, the elimination of Section 230 could discourage generative AI companies from implementing said guardrails in the first place; a countermove that would enable users to prompt LLMs with malicious queries to bait out unlawful responses subject to litigation. In other words, plaintiffs could transform ChatGPT into their very own personal perpetual litigation machine.
And as Perault rightfully warns:
“If a company that deploys an LLM can be dragged into lengthy, costly litigation any time a user prompts the tool to generate text that creates legal risk, companies will narrow the scope and scale of deployment dramatically. Without Section 230 protection, the risk is vast: Platforms using LLMs would be subject to a wide array of suits under federal and state law. Section 230 was designed to allow internet companies to offer uniform products throughout the country, rather than needing to offer a different search engine in Texas and New York or a different social media app in California and Florida. In the absence of liability protections, platforms seeking to deploy LLMs would face a compliance minefield, potentially requiring them to alter their products on a state-by-state basis or even pull them out of certain states entirely…
…The result would be to limit expression—platforms seeking to limit legal risk will inevitably censor legitimate speech as well. Historically, limits on expression have frustrated both liberals and conservatives, with those on the left concerned that censorship disproportionately harms marginalized communities, and those on the right concerned that censorship disproportionately restricts conservative viewpoints.
The risk of liability could also impact competition in the LLM market. Because smaller companies lack the resources to bear legal costs like Google and Microsoft may, it is reasonable to assume that this risk would reduce startup activity.”
Hence, regardless of how we feel about Section 230’s applicability to AI, we will be forced to reckon with the latest iteration of Masnick’s Impossibility Theorem: there is no content moderation system that can meet the needs of all users. The lack of limitations on human awfulness mirrors the constant challenge that social media companies encounter with content moderation. The question is whether LLMs can improve what social media cannot.
With the rise of ChatGPT over the past few months, the inevitable moral panics have begun. We’ve seen a bunch of people freaking out about how ChatGPT will be used by students to do their homework, how it will replace certain jobs, and other claims. Most of these are totally overblown. While some cooler heads have prevailed, and argued (correctly) that schools need to learn to teach with ChatGPT, rather than against it, the screaming about ChatGPT in schools is likely to continue.
To help try to cut off some of that, OpenAI (the makers of ChatGPT) have announced a classification tool that will seek to tell you if something was written by an AI or a human.
We’ve trained a classifier to distinguish between text written by a human and text written by AIs from a variety of providers. While it is impossible to reliably detect all AI-written text, we believe good classifiers can inform mitigations for false claims that AI-generated text was written by a human: for example, running automated misinformation campaigns, using AI tools for academic dishonesty, and positioning an AI chatbot as a human.
And, to some extent, that’s great. Using the tech to deal with the problems created by that tech seems like a good start.
But… human nature raises questions about how this tool will be abused. OpenAI is pretty explicit that the tool is not that reliable:
Our classifier is not fully reliable. In our evaluations on a “challenge set” of English texts, our classifier correctly identifies 26% of AI-written text (true positives) as “likely AI-written,” while incorrectly labeling human-written text as AI-written 9% of the time (false positives). Our classifier’s reliability typically improves as the length of the input text increases. Compared to our previously released classifier, this new classifier is significantly more reliable on text from more recent AI systems.
That… is an extraordinarily high level of both Type I and Type II errors. And that’s likely to create real problems. Because no matter how much you say “our classifier is not fully reliable” human nature says that people are going to treat the output as meaningful. This is the nature of anything that kicks out some sort of answer, it’s hard for humans to wrap their heads around the spectrum of possible actual results. If the computer spits out a “this is possibly” AI generated, or even an “unclear if” (the semi-neutral rating the classifier produces), it’s still going to cause people (teachers especially) to doubt the students.
And, yet, it’s going to be wrong an awful lot.
That seems incredibly risky. We’ve seen this in other areas as well. When computer algorithms are used to recommend criminal sentencing, judges tend to rely on the output as somehow “scientific” even though it’s often bullshit.
I appreciate that OpenAI is trying to help provide the tools to respond to the concerns that some people (teachers and parents, mainly) are raising, but I worry about the backlash in the other direction: the over reliance on this highly unreliable technology from the other end. I mean, we already went through this nonsense with existing plagiarism checking tools, which also run into problems with false positives that can have huge impacts on people’s lives.
That’s not to say there’s no place for this kind of technology, but it’s inevitable that teachers are going to rely on this beyond the level of reliability the tool provides.
Instead, one hopes that schools start figuring out how to use the technology productively. The NY Times article linked above has some good examples:
Cherie Shields, a high school English teacher in Oregon, told me that she had recently assigned students in one of her classes to use ChatGPT to create outlines for their essays comparing and contrasting two 19th-century short stories that touch on themes of gender and mental health: “The Story of an Hour,” by Kate Chopin, and “The Yellow Wallpaper,” by Charlotte Perkins Gilman. Once the outlines were generated, her students put their laptops away and wrote their essays longhand.
The process, she said, had not only deepened students’ understanding of the stories. It had also taught them about interacting with A.I. models, and how to coax a helpful response out of one.
“They have to understand, ‘I need this to produce an outline about X, Y and Z,’ and they have to think very carefully about it,” Ms. Shields said. “And if they don’t get the result that they want, they can always revise it.”
Over on Mastodon, I saw a professor explain how he is using ChatGPT: asking his students to create a prompt to generate an essay about the subject they’re studying, and then having them edit, correct, and rewrite the essay. They would then have to turn in their initial prompt, the initial output, and their revision. I actually think this is a more powerful learning tool than having someone just write an essay in the first place. I know that I learn a subject best when I’m forced to teach it to others (despite taking multiple levels of statistics in college, I didn’t fully feel I understood statistics until I had to teach a freshman stats class, and had to answer student questions all the time). ChatGPT presents a way of making students the “teacher” in this kind of manner, forcing them to more fully understand the issues, and even to correct ChatGPT when it gets stuff wrong.
All of that seems like a more valuable approach to education with AI beyond a semi-unreliable tool to try to “catch” AI-generate text.
Oh, and in case you’re wondering, I ran this article through OpenAI’s classifier and it said:
The classifier considers the text to be very unlikely AI-generated.
Phew. But, knowing how unreliable it is, who can really say?
Every amicus brief the Copia Institute has filed has been important. But the brief filed today is one where all the marbles are at stake. Up before the Supreme Court is Gonzalez v. Google, a case that puts Section 230 squarely in the sights of the Court, including its justices who have previously expressed serious misunderstandings about the operation and merit of the law.
As we wrote in this brief, the Internet depends on Section 230 remaining the intentionally broad law it was drafted to be, applying to all sorts of platforms and services that make the Internet work. On this brief the Copia Institute was joined by Engine Advocacy, speaking on behalf of the startup community, which depends on Section 230 to build companies able to provide online services, and Chris Riley, an individual person running a Mastodon server who most definitely needs Section 230 to make it possible for him to provide that Twitter alternative to other people. There seems to be this pervasive misconception that the Internet begins and ends with the platforms and services provided by “big tech” companies like Google. In reality, the provision of platform services is a profoundly human endeavor that needs protecting in order to be sustained, and we wrote this brief to highlight how personal Section 230’s protection really is.
Because ultimately without Section 230 every provider would be in jeopardy every time they helped facilitate online speech and every time they moderated it, even though both activities are what the Internet-using public needs platforms and services to do, even though they are what Congress intended to encourage platforms and services to do, and even though the First Amendment gives them the right to do them. Section 230 is what makes it possible at a practical level for them to them by taking away the risk of liability arising from how they do.
This case risks curtailing that critical statutory protection by inventing the notion pressed by the plaintiffs that if a platform uses an algorithmic tool to serve curated content, it somehow amounts to having created that content, which would put the activity beyond the protection of Section 230 as it only applies to when platforms intermediate content created by others and not content created by themselves. But this argument reflects a dubious read of the statute, and one that would largely obviate Section 230’s protection altogether by allowing liability to accrue as a result of some quality in the content created by another, which is exactly what Section 230 is designed to forestall. As we explained to the Court in detail, the idea that algorithmic serving of third party content could somehow void a platform’s Section 230 protection is an argument that had been cogently rejected by the Second Circuit and should similarly be rejected here.
Oral argument is scheduled for February 21. While it is possible that the Supreme Court could take onboard all the arguments being brought by Google and the constellation of amici supporting its position, and then articulate a clear defense of Section 230 platform operators could take back to any other court questioning in their statutory protection, it would be a good result if the Supreme Court simply rejected this particular theory pressing for artificial limits to Section 230 that are not in the statute or supported by the facially obvious policy values Section 230 was supposed to advance. Just so long as the Internet and the platforms that make up it can live on to fight another day we can call it a win. Because a decision in favor of the plaintiffs curtailing Section 230 would be an enormous loss to anyone depending on the Internet to provide them any sort of benefit. Or, in other words, everyone.
In early December 2022, a former Israeli Minister of Defense and Chief of Staff of the Israel Defense Forces, three other retired Israeli generals, a former Commissioner of the Israeli Police, and a former head of the Mossad’s Intelligence Directorate filed an amicus brief before the U.S. Supreme Court in Gonzalez v. Google arguing that Internet platforms should be civilly liable for third party content that encourages terrorist activity. In their filing, they claimed that the wave of terror in Israel in 2015–2016 “became known as the ‘Facebook intifada’ and the #stab! Campaign due to the essential role social media played in inciting the perpetrators to attack civilians.” The Anti-Defamation League also filed a brief in the case, similarly arguing that Internet platforms should have legal accountability for violence against Jewish Americans and other vulnerable communities encouraged by these platforms’ recommendation engines. So, too, the Zionist Organization of America asserted that Internet platforms should not be immune from liability “when they target specific users and recommend and direct them to new content that helps fan the flames of hatred and violence against the Jewish community.”
There is no doubt that Internet platforms are used to disseminate antisemitic content. But these briefs fail to recognize that these same platforms greatly foster Jewish community and religious activity in the United States and throughout the Diaspora; and that the legal interpretations these briefs advocate could drastically diminish this activity.
In Gonzalez, the U.S. Supreme Court will consider for the first time the scope of the safe harbor provided by Section 230 of the Communications Decency Act, which limits the liability of an Internet platform for content uploaded by third parties. Since Section 230’s enactment in 1996, the lower courts have interpreted it broadly. Some politicians and interest groups argue that these interpretations are overly broad and have disincentivized Internet companies from eliminating hate speech and disinformation from their sites. Free speech advocates, on the other hand, contend that the courts have correctly applied Section 230 in a manner that enables platforms to allow any speaker to reach a global audience at no, or minimal, cost, without prior vetting or filtering. In Gonzalez, the Supreme Court could uphold the broad interpretation upon which Internet companies have relied for the past quarter century, or it instead could narrow the scope of the safe harbor and disrupt the existing business models of the Open Internet.
Increasing the liability of the platforms for the third-party content would force the platforms to act as gatekeepers; to reduce their exposure to ruinous damages, many platforms would permit dissemination only of paid or pre-approved content. This could adversely affect a wide range of online activity, including the rich Jewish life facilitated by the Internet.
Worship. Even before the Covid-19 pandemic, Jewish congregations had begun to experiment with the live streaming of religious services. Live streaming and video conferencing of services increased dramatically with the onset of the pandemic, and many congregations now use hybrid models. While Orthodox congregations will not stream their services on the Sabbath and the holidays, they will stream the daily morning, afternoon, and evening services. Less strict denominations not concerned about the use of electricity will also stream services on the Sabbath and holidays. (For example, this past Rosh HaShanah, I witnessed the blowing of the shofar at B’nai Jeshuran Congregation in New York from my hotel room in Geneva, Switzerland.) Indeed, the use of video conferencing and streaming technologies have led to extensive rabbinic debate whether people participating in a service via Zoom counted towards the “minyan” or quorum of ten participants. The Rabbinical Assembly of the Conservative Movement issued a 50-page legal opinion on the subject.
The availability of remote attendance has led to growing participation in daily services and the strengthening of ties to Judaism. One now can routinely join shiva minyans at the houses of mourners via Zoom or other video-conferencing platforms, enabling mourners to be joined by family and around the world. The same is true with other life-cycle events, such as brises (ritual circumcisions) and weddings.
Education. Internet platforms provide myriad channels for formal and informal Jewish education. During the pandemic, Jewish institutions of learning at all levels migrated online using platforms such as Zoom or Webex. Adult education programs on Jewish topics by synagogues, universities, and other organizations are now offered online. Lectures are live streamed and archived on YouTube. Hundreds of rabbis from around the world offer “daf yomi” or the daily study of a page of Talmud via Internet platforms.
Culture and Community-Engagement. Social media platforms such as YouTube host vast quantities of Jewish cultural material, including videos of performances of songs and dances. Around holidays, groups such as Six13 and the Maccabeats release their latest holiday-themed recordings. Synagogues and other Jewish organizations use platforms like Facebook and Zoom for cultural events, book groups and professional discussion forums for rabbis, cantors, and teachers. Hadassah Magazine in an article entitled A (Facebook) Group for Every Jewish Interest reviewed some of the over 1,000 Facebook groups with “Jewish” or “Jews” in their names.
In short, Internet platforms allow Jews in the Diaspora to practice their faith and strengthen their identity. Moreover, U.S.-based platforms heavily support all aspects of political, economic, cultural, and personal life in Israel. Indeed, Israelis are the world leaders in social media use, with 77 percent of adults using social platforms such as Facebook, Instagram, and WhatsApp. Imposing greater liability on U.S. platforms for third party content could endanger these positive uses by increasing their cost and reducing their spontaneity.
Proponents of narrowing the Section 230 safe harbor, including the generals and organizations mentioned above, may contend that rather than wholesale changes to the application of Section 230, they merely want increased liability for the use of algorithms recommending content. But virtually all social media sites use recommendation algorithms; the amount of content available on the Internet is so enormous that all search engines and sites hosting content use algorithms to determine what content to offer users.
The Solicitor General of the United States in its brief in the Gonzalez case tried to draw a distinction between the use of algorithms by search engines to select content in response to a user’s query and the use of algorithms by a social media platform to supply a user with content by an automatically generated feed. This is a distinction without a difference. The search engine algorithm considers searches the user has previously conducted in determining what search results to present the user in response to the particular query he is now making; the platform considers the user’s prior activity in determining what content to present the user in her feed. In both cases, the user’s prior activity influences the algorithm.
Furthermore, even if there were a difference between search engine results and feeds, feeds are extremely beneficial to the user and society at large in most cases. The feed provides the user with more of the content she wants to see, and usually that content is not problematic in any way. To be sure, a social media platform might feed additional antisemitic content to a user who spends some time on the platform viewing antisemitic content. By the same token, the platform would feed Jewish educational material to a user who spends time on the platform viewing Jewish educational material. Platforms should not be forced to abandon feeds, with all the resulting user benefits, because on occasion the feeds may have harmful impacts.
Finally, three quick responses to the suggestion that platforms could easily remove access to antisemitic content without changing their business models in a manner that ultimately restricts access to legitimate content. First, display of antisemitic symbols and content might be necessary for educational purposes, such as to teach about the Holocaust — but Internet companies have difficulty accurately making such content moderation distinctions, particularly at scale. Second, there is profound disagreement about when criticism of Israeli government policies towards the West Bank and Gaza Strip constitutes antisemitism. Here, too, Internet companies have trouble getting the nuance right. Third, even if the social media platforms could draw appropriate lines with respect to antisemitic material, changes to Section 230 would still lead to liability for other problematic content, and the platforms would still need to change their business models, to the detriment of Jewish activity online.
A decision in the Gonzalez case is expected by the end of June 2023.
Jonathan Band is a copyright and internet lawyer based in Maryland. This article was reposted with permission.
So, plenty of Supreme Court watchers and Section 230 experts all knew that this term was going to be a big one for Section 230… it’s just that we all expected the main issue to be around the Netchoice cases regarding Florida and Texas’s social media laws (those cases will likely still get to SCOTUS later in the term). There were also a few other possible Section 230 cases that I thought SCOTUS might take on, but still, the Court surprised me by agreeing to hear two slightly weird Section 230 cases. The cases are Gonzalez v. Google and Twitter v. Taamneh.
There are a bunch of similar cases, many of which were filed by two law firms together, 1-800-LAW-FIRM (really) and Excolo Law. Those two firms have been trying to claim that anyone injured by a terrorist group should be able to sue internet companies because those terrorist groups happened to use those social media sites. Technically, they’re arguing “material support for terrorism,” but the whole concept seems obviously ridiculous. It’s the equivalent of the family of a victim of ISIS suing Toyota after finding out that some ISIS members drove Toyotas.
Anyway, we’ve been writing about a bunch of these cases, including both of the cases at issue here (which were joined at the hip by the 9th Circuit). Most of them get tossed out pretty quickly, as the court recognizes just how disconnected the social media companies are from the underlying harm. But one of the reasons they seem to have filed so many such cases all around the country was to try to set up some kind of circuit split to interest the Supreme Court.
The first case (Gonzalez) dealt with ISIS terrorist attacks in Paris in 2015. The 9th Circuit rejected the claim that Google provided material support to terrorists because ISIS posted some videos to YouTube. To try to get around the obvious 230 issues, Gonzalez argued that YouTube recommended some of those videos via the algorithm, and those recommendations should not be covered by 230. The second case, Taamneh, was… weird. It has a somewhat similar fact pattern, but dealt with the family of someone who was killed by an ISIS attack at a nightclub in Istanbul in 2017.
The 9th Circuit tossed out the Gonzalez case, saying that 230 made the company immune even for recommended content (which is the correct outcome) but allowed the Taamneh case to move forward, for reasons that had nothing to do with Section 230. In Taamneh, the district court initially dismissed the case entirely without even getting to the Section 230 issue by noting that Taamneh didn’t even file a plausible aiding-and-abetting claim. The 9th Circuit disagreed, said that there was enough in the complaint to plead aiding-and-abetting, and sent it back to the district court (which could then, in all likelihood, dismiss under Section 230). Oddly (and unfortunately) some of the judges in that ruling issued concurrences which meandered aimlessly, talking about how Section 230 had gone too far and needed to be trimmed back.
Gonzalez appealed the issue regarding 230 and algorithmic promotion of content, while Twitter appealed the aiding and abetting ruling (noting that every other court to try similar cases found no aiding and abetting).
Either way, the Supreme Court is taking up both cases and… it might get messy. Technically, the question the Supreme Court is asked to answer in the Gonzalez case is:
Whether Section 230(c)(1) of the Communications Decency Act immunizes interactive computer services when they make targeted recommendations of information provided by another information content provider, or only limits the liability of interactive computer services when they engage in traditional editorial functions (such as deciding whether to display or withdraw) with regard to such information.
Basically: can we wipe out Section 230’s key liability protections for any content recommended? This would be problematic. The whole point of Section 230 is to put the liability on the proper party: the one actually speaking. Making sites liable for recommendations creates all of the same problems that making them liable for hosting would — specifically, requiring them to take on liability for content they couldn’t possibly thoroughly vet before recommending it. A ruling in favor of Gonzalez would create huge problems for anyone offering search on any website, because a “bad” content recommendation could lead to liability, not for the actual content provider, but for the search engine.
That can’t be the law, because that would make search next to impossible.
For what it’s worth, there were some other dangerously odd parts of the 9th Circuit’s Gonzalez rulings regarding Section 230 that are ripe for problematic future interpretation, but those parts appear not to have been included in the cert petition.
In Taamneh, the question is focused on the aiding and abetting question, but ties into Section 230, because it asks if you can hold a website liable for aiding and abetting if they try to remove terrorist content but a plaintiff argues they could have been more aggressive in weeding out such content. There’s also a second question of whether or not you can hold a website liable for an “act of intentional terrorism” when the actual act of terrorism had nothing whatsoever to do with the website, and was conducted off of the website entirely.
(1) Whether a defendant that provides generic, widely available services to all its numerous users and “regularly” works to detect and prevent terrorists from using those services “knowingly” provided substantial assistance under 18 U.S.C. § 2333 merely because it allegedly could have taken more “meaningful” or “aggressive” action to prevent such use; and (2) whether a defendant whose generic, widely available services were not used in connection with the specific “act of international terrorism” that injured the plaintiff may be liable for aiding and abetting under Section 2333.
These cases should worry everyone, especially if you like things like searching online. My biggest fear, honestly, is that this Supreme Court (as it’s been known to do) tries to split the baby (which, let us remember, kills the baby) and says that Section 230 doesn’t apply to recommended content, but that the websites still win because the things on the website are so far disconnected from the actual terrorist acts.
That really feels like the kind of solution that the Roberts court might like, thinking that it’s super clever when really it’s just dangerously confused. It would open up a huge pandora’s box of problems, leading to all sorts of lawsuits regarding any kind of recommended content, including search, recommendation algorithms, your social media feeds, and more.
A good ruling (if such a thing is possible) would be a clear statement that of course Section 230 protects algorithmically rated content, because Section 230 is about properly putting liability on the creator of the content and not the intermediary. But we know that Justices Thomas and Alito are just itching to destroy 230, so we’re already down two Justices to start.
Of course, given that this court is also likely to take up the NetChoice cases later this term, it is entirely possible that next year the Supreme Court may rules that (1) websites are liable for failing to remove certain content (in these two cases) and(2) websites can be forced to carry all content.
It’ll be a blast figuring out how to make all that work. Though, some of us will probably have to do that figuring out off the internet, since it’s not clear how the internet will actually work at that point.
As you may recall, back during the Trump administration, after a bunch of kids on TikTok trolled Trump into believing one of his campaign rallies would be massively attended (which it was not), Trump decided to take out his anger on TikTok by issuing an almost certainly unconstitutional executive order demanding that TikTok’s owner, the Chinese firm ByteDance, sell TikTok to an American company. While a few potential buyers lined up to pick up the increasingly popular social media company on the cheap (due to the forced sale nature of it), White House insiders revealed that they would only approve the sale if it went to a friend of Donald Trump’s (this, of course, is corrupt nonsense, but hey, no one cares about that any more). That left precious few options, as Trump wouldn’t approve the sale to the few companies that actually wanted to buy the whole thing outright: namely Microsoft and Walmart.
In the end, Trump wanted the company to go to his buddy Larry Ellison’s Oracle. Of course, there was a problem: Oracle had no use for TikTok as a subsidiary. Oracle does enterprise stuff, not social media. But, what Oracle does have is a cloud hosting offering that is way down the list behind industry leaders like Amazon, Microsoft, IBM, Google and others. So, Oracle and the Trump administration cooked up… a hosting deal for Oracle.
Basically, Oracle would get TikTok’s US hosting business with some vague promises of protecting data privacy, while Trump would get to help out a friend (Ellison) while pretending he’d actually accomplished something (even though it wasn’t at all what he initially demanded). Of course, this was all about posturing and headlines, so not much came of the deal for a while.
But, with new (somewhat questionable) claims about US TikTok data being accessible to ByteDance employees making news, the company apparently (two years later) has started to make good on the deal and in June announced that all of its US data was routed to Oracle.
It’s not exactly clear what this means in practice — and we’ll remind folks that there were reports last year claiming that Oracle had Chinese law enforcement customers, which raised at least some questions about its actual commitment to protecting data from the Chinese government. Also notable: Oracle has spent years gleefully trying to undermine basically all content moderation by funding groups to advocate against Section 230. Oh, and I guess we should mention, that for all the claims of TikTok being controlled by the Chinese government, remember that Oracle got its start… as a CIA project. There is something richly ironic in the idea that Oracle is somehow a trustworthy partner here.
Given all that — what exactly does it mean for Oracle to be “auditing” TikTok’s algorithms and content moderation? Given that the company doesn’t have the best track record on privacy and has worked to undermine content moderation for years now, the whole thing is… just kinda strange. Oracle’s explanation is not very clear at all:
The reviews give Oracle visibility into how TikTok’s algorithms surfaces content “to ensure that outcomes are in line with expectations and that the models have not been manipulated in any way,” the spokesperson said.
I mean, what does “manipulated” even mean in that sentence? Of course they’re manipulated. Someone wrote the algorithm. If they mean “not manipulated to promote Chinese propaganda” or “not manipulated to suppress anti-Chinese content” then… maybe say that. Because “manipulation” on its own doesn’t mean anything reasonable here.
There is nothing in Oracle’s history or experience that suggests the company has any useful insight into how TikTok handles recommendations or content moderation. There are plenty of reasons to think that Oracle might actually be problematic in this role.
The whole setup seems quite strange, and really feels like everyone just sort of making it up as they go along. TikTok needs some sort of US oversight to appease people who are freaked out that a Chinese-owned social media company is successful in the US, and Oracle was right there to say it would do it, in exchange for a lucrative hosting deal for its lagging cloud offering. This also feels vaguely similar to how the US has been accusing Chinese firms like Huawei and ZTE of using their tech to snoop on people… when that’s actually exactly what the US government has been doing via Cisco for years.
Also, what kind of precedent does this set? Will we be okay if other countries demand that their own favored companies have to audit US firms’ algorithms and content moderation practices? Because… that is going to create quite a mess.
