Generative AI Algorithms: The Fine Line Between Speech and Section 230 Immunity

Generative AI Algorithms: The Fine Line Between Speech and Section 230 Immunity

 By Hannah G. Babinski

ABSTRACT

Russian-American writer and philosopher Ayn Rand once observed, “No speech is ever considered, but only the speaker. It’s so much easier to pass judgment on a man than on an idea.”[1] But what if the speaker is not a man, woman, or a human at all? Concepts of speech and identities of speakers have been the focal points of various court cases and debates in recent years. The Supreme Court and various district courts have faced complex and first-of-their-kind questions concerning emerging technologies, namely algorithms and recommendations, and contemplated whether their outputs constitute speech on behalf of an Internet service provider (“Internet platform”) that would not be covered by Section 230 of the Communications Decency Act (“Section 230”).  In this piece, I will examine some of the issues arising from the questions posed by Justice Gorsuch in Gonzalez v. Google, LLC, namely whether generative AI algorithms and their relative outputs constitute speech that is not immunized under Section 230. I will provide an overview of the technology behind generative AI algorithms and then examine the statutory language and interpretation of Section 230, applying that language and interpretive case law to generative AI. Finally, I will provide demonstrative comparisons between generative AI technology and human content creation and foundational Copyright Law concepts to illustrate how generative AI technologies and algorithmic outputs are akin to unique, standalone products that extend beyond the protections of Section 230.

 

Continue Reading 

Section 230 and Radicalization Scapegoating

Section 230 and Radicalization Scapegoating

By Hannah G. Babinski, Class of 2024

Standing as one of the few provisions of the Communications Decency Act of 1996 yet to be invalidated by the Court as unconstitutional, 47 U.S.C. § 230 (“Section 230”) has repeatedly been at the center of controversy since its enactment. As the modern world continues to become further dependent on online, electronic communication, such controversy is likely to only grow. Section 230 insulates interactive computer services—think social media websites, chat-boards, and any other website that enables a third-party user of the website to upload a post, text, video, or other medium of expression—from liability stemming from content uploaded to the website by third-party users, even where interactive computer services engage in good-faith content moderation. In this regard, the provision effectively serves to classify the third parties, and not the host website, as the speakers or publishers of content.

Though Section 230 has been instrumental in the development of the internet at large, by preventing needless and substantial litigation and establishing a sense of accountability for individual users in tort generally, the limited language of Section 230 has resulted in several issues of interpretation concerning the line between what actions, specifically content moderation, constitute speech on behalf of the interactive computer service provider and what actions do not. Over the course of the last five years, courts have examined in particular whether algorithms created by and incorporated into the host websites are speech and, thus, unprotected by Section 230.

In Force v. Facebook, Inc., the Court of Appeals for the Second Circuit addressed the question of algorithms as speech in the context a Facebook algorithm that directed radicalized content and other pages openly maintained and associated with the terrorist organization Hamas, a Palestinian radical Islamist organization, to the personalized newsfeeds of several individuals, who then went on to attack five Americans in Israel between 2014 and 2016.[1]

Though the majority opinion ultimately concluded that the algorithm was protected by Section 230 immunity, Chief Judge Katzmann dissented with a well-written and thorough argument against applying Section 230 immunity to such a case. Though I reserve my opinion concerning whether I necessarily agree or disagree with the dissent in Force v. Facebook, Inc., Katzmann verbalizes the key concern with Section 230 as it applies to social media as a whole, stating:

By surfacing ideas that were previously deemed too radical to take seriously, social media mainstreams them, which studies show makes people “much more open” to those concepts. . . . The sites are not entirely to blame, of course—they would not have such success without humans willing to generate and to view extreme content. Providers are also tweaking the algorithms to reduce their pull toward hate speech and other inflammatory material. . . . Yet the dangers of social media, in its current form, are palpable.[2]

This statement goes to the heart of the controversy surrounding not only algorithms, but exposure to harmful or radicalizing content on the internet generally, which is exacerbated by the advent and use of social media platforms; with the expansive and uninhibited nature of the internet ecosystem and social media websites enabling and even facilitating the connection between certain individuals with a proclivity for indoctrination and individuals disseminating radicalized content absent the traditional restrictions of time, language or national borders, it is only natural that greater radicalization has resulted. Does this mean that we, as a society, should hinder communication in order to prevent radicalization?

Proponents of dismantling Section 230 and casting the onus on interactive computer service providers to engage in more rigorous substantive moderation efforts would answer that question in the affirmative. However, rather than waging war on the proverbial middleman and laying blame on communication outlets, we should instead concentrate our efforts on the question, acknowledged by Katzmann, of why humans seem more willing to generate and consume extremist content in the modern age. We, as a society, should take responsibility for the increase in radicalized content and vulnerabilities that are resulting in higher individual susceptibility to radicalization, tackling what inspires the speaker as opposed to the tool of speech.

According to findings of the Central Intelligence Agency (“CIA”) and affirmed by the Federal Bureau of Investigation (“FBI”), certain vulnerabilities are almost always present in any violent extremist, regardless of ideology or affiliation; these vulnerabilities include “feeling alone or lacking meaning and purpose in life, being emotionally upset after a stressful event, disagreeing with government policy, not feeling valued or appreciated by society, believing they have limited chances to succeed, [and] feeling hatred toward certain types of people.”[3] As these vulnerabilities are perpetuated by repeated societal and social failures, the number of susceptible individuals will continue to climb.

What’s more, these predispositions are not novel to the age of social media. Undoubtedly, throughout history, we have seen the proliferation of dangerous cults and ideological organizations that radicalize traditional beliefs, targeting the dejected and the isolated in society. For example, political organizations like the National Socialist German Workers’ Party, more infamously known as the NAZI party; Christianity-based cults and hate organizations like the People’s Temple, Children of God, Branch Davidians, and the Klu Klux Klan; and Buddhist-inspired terrorism groups like Aum Shinrikyo have four things in common: 1) they radicalized impressionable individuals, many of whom experienced some of the vulnerabilities cited above, 2) they brought abuse/harm/death to members, 3) they facilitated and encouraged abuse/harm/death to nonmembers, and 4) they reached popularity and obtained their initial members without the help of algorithmic recommendations and social media exposure.

The point is that social media is not to blame for radicalization. Facebook and YouTube’s code-based algorithms that serve to connect individuals with similar interests on social networking sites or organize content based on individualized past video consumption are not to blame for terrorism. We are.

[1] Force v. Facebook, Inc., 934 F. 3d 53 (2d Cir. 2019).

[2] Id.

[3] Cathy Cassasta, Why Do People Become Extremists?, Healthline (updated Sept. 18, 2017), https://www.healthline.com/health-news/why-do-people-become-extremists (last visited Feb. 26, 2023).