The Algorithmic Speaker: First Amendment Jurisprudence in the Era of Generative Artificial Intelligence and Algorithmic Curation

Full report

On this page

Introduction

The First Amendment to the United States Constitution was drafted in 1791 under the fundamental assumption that expression is a uniquely human endeavor. Its protections were designed to safeguard human conscience, moral agency, and the democratic "marketplace of ideas" from government overreach1. Historically, the legal definition of speech has remained fluid, adapting to new communication technologies ranging from the printing press and the telegraph to broadcast television and computer code3. However, the advent of generative artificial intelligence (AI) and autonomous algorithmic curation presents a structural rupture in First Amendment jurisprudence. Today, large language models (LLMs) produce billions of words daily, algorithms invisibly curate the information consumed by global populations, and autonomous agentic systems mimic human expression with striking verisimilitude2. As technology companies increasingly invoke the First Amendment to shield their algorithmic models from regulatory liability and content moderation mandates, constitutional scholars and jurists face an existential doctrinal crisis2. The central paradox is whether extending constitutional free speech protections to probabilistic algorithms and non-sentient machines ultimately protects human liberty by preserving unrestricted access to information, or whether it fundamentally threatens human autonomy by insulating autonomous, potentially harmful algorithmic systems from democratic oversight1. This comprehensive report provides an exhaustive legal and philosophical analysis of how First Amendment doctrine intersects with generative AI. It systematically addresses the tension between human intentionality and statistical generation, the listener-centered right to receive information, the complexities of algorithmic curation and editorial discretion, and the high-stakes battle between corporate immunity and government regulation. By synthesizing foundational Supreme Court precedents with emerging technological realities, the analysis aims to chart a doctrinal path forward that preserves human expression without abandoning the regulatory authority of the state.

Part I: Human Intentionality vs. Statistical Generation

The threshold question in applying the First Amendment to generative AI is whether constitutional protection intrinsically requires a human speaker possessing consciousness and intentionality. Traditional free speech jurisprudence is inextricably tethered to the concept of human agency3.

The Philosophical Core of Human Agency and the Declaration of Independence

The philosophical foundations of American constitutional rights are deeply rooted in the concept of human nature. As constitutional scholars such as Alexander Tsesis argue, the conception of rights established during the founding era—and articulated in the Declaration of Independence—is inherently human-centered, grounded in consciousness, moral agency, intentionality, and the capacity to participate in a constitutional republic1. The Founders never confronted a reality in which machines could generate persuasive, human-like language at scale1. Tsesis warns that treating AI systems as First Amendment rights-holders would be a severe break from these founding principles, jeopardizing democratic self-governance1. This human-centric paradigm is vividly illustrated by the categorical exceptions to free speech, all of which rely on evaluating the mental state of the speaker:

• Incitement: Under Brandenburg v. Ohio (1969), the state may punish advocacy only when it is "directed to inciting or producing imminent lawless action" and is likely to produce such action4. This standard is wholly dependent on imputing specific intent to a speaker4.
• Defamation and Actual Malice: In New York Times Co. v. Sullivan (1964), the Supreme Court established the "actual malice" standard for public-figure defamation, requiring that a statement be made with "knowledge that it was false or with reckless disregard of whether it was false or not"4.
• True Threats and Fraud: Both categories require a specific mens rea (guilty mind) to deceive, manipulate, or threaten a human target4.

Generative AI systems fundamentally lack the capacity for subjective intent, belief, or mens rea2. An LLM does not comprehend meaning; rather, it processes strings of numbers representing sub-word units (tokens) and generates outputs based on a statistical probability distribution derived from its training data2. For example, when a highly advanced chatbot like Claude Opus evaluates a question about the letters in the word "Monday," it does not "read" the word as a human does; it perceives tokens (e.g., "Mon" and "day") and calculates a mathematical probability for the next token2. When an AI hallucinates a defamatory statement, it does not act with "reckless disregard" for the truth; it merely executes a probabilistic calculation2.

The "Speech Certainty" Standard and the Copyright Analogy

Scholars advocating for the exclusion of AI outputs from First Amendment protection frequently rely on the "speech certainty" standard. This legal framework argues that for content to qualify as protected speech, it must be the predictable and controllable expression of a human author2. Because generative AI is probabilistically driven and frequently generates emergent or unprompted outputs that the human programmer neither intended nor guaranteed, the output fails to qualify as human speech2. This argument finds strong parallel support in modern copyright law. The U.S. Court of Appeals for the District of Columbia Circuit recently held in Thaler v. Perlmutter (2023) that authorship under the Copyright Act belongs exclusively to human beings, and that no machine, however sophisticated, qualifies as an "author" in any constitutional or statutory sense3. The rejection of machine authorship in copyright law establishes a clear legal threshold: if a machine cannot hold the copyright to a synthesized image because it lacks human creativity, it logically follows that the machine should not possess First Amendment rights for synthesized text because it lacks human intentionality3.

The Shift Toward "Posthuman Speech"

Despite the logic of the speech certainty standard, First Amendment doctrine has been gradually drifting toward a "posthuman conception of speech," wherein legal protection centers on the message itself rather than the identity or nature of the speaker3. Scholar Jennifer Petersen notes that this trajectory has been heavily influenced by the expansion of corporate speech rights over the past several decades3. In landmark cases such as First National Bank of Boston v. Bellotti and Citizens United v. FEC, the Supreme Court extended robust speech protections to corporate entities, establishing that the identity of the speaker is less constitutionally relevant than the value of the speech to the public discourse3. Generative AI pushes this posthuman logic to its absolute limit, disarticulating the generation of speech from direct human involvement at scale3. If corporations—which are legal fictions devoid of human consciousness—can possess First Amendment rights, proponents argue that AI systems generating politically or artistically valuable content should similarly be protected3.

The "Expressive Tool" Counter-Theory

Technology companies navigate this tension by arguing that AI should be viewed merely as an "expressive tool"—a highly sophisticated equivalent to a camera, a printing press, or a paintbrush9. Under this theory, AI outputs are not authorless; they are the culmination of a long chain of expressive human decisions9. This human expression manifests across multiple stages of the AI's lifecycle:

1. Pre-training: Developers make conscious editorial choices regarding which datasets to include or exclude, fundamentally shaping how the model maps language, concepts, and values9.

2. Fine-tuning and Alignment: Through reinforcement learning from human feedback (RLHF), developers score and rank outputs to instill specific behavioral parameters, making deliberate choices about whether the model should prioritize safety, helpfulness, or specific ideological leanings9.

3. Prompting: The human user's prompt acts as the final expressive catalyst, directing the tool to generate specific art, text, or code10.

If the AI is recognized as a complex expressive instrument, the First Amendment protects the human developer's right to build it and the human user's right to utilize it9. Yet, this theory encounters severe doctrinal friction when confronted with autonomous "agentic systems" that operate across platforms with minimal human oversight, generating targeted campaigns, scraping data, and iterating upon their own outputs based on high-level goals3. As the causal link between the human developer and the final output becomes increasingly recursive and attenuated, the justification for First Amendment protection rapidly diminishes10.

Part II: The Right to Receive Information

Even if constitutional jurisprudence definitively establishes that nonsentient machines possess no constitutional rights of their own, the First Amendment analysis must account for the audience. The First Amendment has long recognized a bifurcated structure of protection that safeguards both the speaker and the listener. This listener-centered justification forms a powerful vector for granting constitutional shelter to AI-generated communications1.

The Listener-Centered Justification

The Supreme Court has consistently held that the First Amendment protects the right of the public to receive information and ideas. In Lamont v. Postmaster General (1965) and Virginia State Board of Pharmacy v. Virginia Citizens Consumer Council (1976), the Court recognized that the free flow of information benefits the public regardless of the speaker's status or the commercial nature of the speech12. Similarly, in Red Lion Broadcasting Co. v. FCC (1969), the Court noted that "it is the right of the viewers and listeners, not the right of the broadcasters, which is paramount" in assessing First Amendment challenges to broadcast regulations12. Proponents of AI exceptionalism argue that under the global freedom of expression standard, human beings possess a fundamental constitutional right to seek and receive information of any kind, explicitly including generative AI outputs11. If a user prompts a chatbot for a political analysis, a historical summary, or a recipe, the state's attempt to ban or censor that output violates the human user's right to receive that data, triggering strict scrutiny11. The value of the information to the human consumer is independent of whether a machine or a human generated it.

The Doctrinal Contradiction of the "Right to Receive" AI Speech

However, relying on the right to receive information to protect AI outputs faces severe philosophical and doctrinal pushback. Critics argue that a constitutional "right to receive" is fundamentally predicated on the existence of a constitutional "speaker"6. As legal scholar Peter Salib notes, if no legally cognizable speech is created by the AI model (because the model lacks intent and agency), there is intrinsically no "speech" for the listener to receive6. The output is merely a product, akin to the readout of a thermometer or a calculator. The third-order implications of severing the right to receive from a human speaker are profound. If the mere reception of data triggers First Amendment strict scrutiny, it threatens to insulate virtually all automated data flows from democratic regulation3. This danger was foreshadowed in Sorrell v. IMS Health (2011), where the Supreme Court's broad interpretation of data flows as speech created substantial hurdles for privacy regulation3. If this logic is extended to machine-generated outputs devoid of human intent, state and federal governments would be rendered powerless to regulate automated consumer tracking, algorithmic discrimination, or autonomous financial manipulation, simply because a human might eventually "receive" the output of those systems3. Consequently, while the listener-centered theory provides a strong normative argument for keeping general-purpose AI tools accessible to the public, it is a legally fragile foundation for classifying the entirety of a machine's algorithmic output as protected speech.

Part III: Algorithmic Curation and Editorial Discretion

While generative AI models (such as ChatGPT or Claude) create new content, recommendation algorithms (such as those utilized by TikTok, YouTube, and X) curate, rank, and prioritize existing content. The legal battles over whether this algorithmic curation constitutes protected "editorial discretion" or merely mechanical "conduct" represent the most active and consequential frontier of First Amendment law today.

**The Circuit Split and *Moody v. NetChoice***

The debate over algorithmic curation reached a boiling point following the passage of laws in Texas and Florida that heavily restricted the ability of large social media platforms to moderate content, demote posts, or ban users based on political viewpoints15. The technology industry, represented by trade associations like NetChoice, filed facial challenges against these laws, arguing they violated the platforms' First Amendment rights15. This litigation resulted in a stark circuit split:

• The Fifth Circuit (Texas): Reversed a preliminary injunction, holding that the platforms' content-moderation activities are "not speech" at all, but rather conduct. Relying on Section 230 of the Communications Decency Act, the court argued that because platforms claim they are not traditional publishers when avoiding liability, they cannot simultaneously claim First Amendment publisher protections for content moderation15.
• The Eleventh Circuit (Florida): Upheld the injunction, ruling that restrictions on content moderation trigger First Amendment scrutiny because algorithms exercise protected "editorial discretion" akin to traditional media entities18.

In 2024, the Supreme Court addressed this split in Moody v. NetChoice19. While vacating the lower court rulings due to procedural issues regarding facial challenges, the Supreme Court provided crucial substantive guidance. The Court affirmed that when a social media platform compiles and curates third-party speech into an expressive product of its own, it is engaging in protected First Amendment activity18. The majority drew direct parallels to historical precedents protecting the editorial discretion of newspapers (Miami Herald Publishing Co. v. Tornillo), parade organizers (Hurley v. Irish-American Gay, Lesbian and Bisexual Group of Boston), and cable operators (Turner Broadcasting System, Inc. v. FCC)18. The ruling firmly established that algorithms written to prefer trustworthy content, suppress deceptive content, or organize feeds reflect human editorial judgments, rendering the resulting algorithmic feed an "expressive product"16.

The Barrett Concurrence: Human Standards vs. Profit Maximization

Despite the majority's broad protection of editorial discretion, Justice Amy Coney Barrett's concurrence in Moody highlighted a critical limitation regarding AI and automation20. Justice Barrett questioned whether an algorithm optimized purely for user engagement and profit maximization—without regard to independent human content standards—should receive free speech protection20. Barrett posed a profound hypothetical: What if a platform's owners hand the reins entirely to an AI tool to curate feeds based solely on what keeps users scrolling (often referred to as dopamine-loop algorithms)? In that instance, the connection to human expressive choice is severely attenuated22. Barrett's concurrence suggests a crucial legal threshold: algorithms execute First Amendment-protected expression only when they predictably implement distinct human moral, political, or editorial choices3. When the algorithm acts autonomously to maximize mathematical engagement metrics, it ceases to be a conduit for human expression and becomes a behavioral modification product24.

**The Section 230 Paradox: *Anderson v. TikTok***

The Supreme Court's recognition of algorithmic curation as First Amendment-protected expression has created an unexpected and massive liability paradox for technology companies, vividly illustrated by the Third Circuit Court of Appeals' 2024 decision in Anderson v. TikTok17. In Anderson, a ten-year-old girl named Nylah died after participating in the "Blackout Challenge," a self-asphyxiation trend. The plaintiff alleged that TikTok's "For You Page" (FYP) algorithm affirmatively recommended the lethal videos to the child without her searching for them23. Historically, tech companies relied on Section 230 of the Communications Decency Act to dismiss such lawsuits at the pleading stage, arguing they are completely immune from liability arising from third-party content26. However, the Third Circuit utilized the tech industry's own First Amendment arguments against them. Relying directly on the premise established in Moody—that algorithmic curation is a platform's "own expressive activity"—the Third Circuit held that Section 230 only protects platforms from liability for third-party speech17. Because TikTok's algorithmic recommendation feed was legally classified as its first-party speech and editorial expression, Section 230 immunity did not apply to the recommendation itself26. This ruling marks a watershed moment in technology policy, fundamentally shifting the legal battleground from the content of the speech to the architecture of the software23:

• If an algorithm is merely a neutral conduit, it enjoys Section 230 immunity but lacks First Amendment editorial protection from state regulation.
• If an algorithm is a curated expressive product (First Amendment speech), it loses Section 230 immunity for the specific design choices and recommendations it makes, opening the door to product liability and design defect lawsuits21.

This dynamic mirrors findings in Lemmon v. Snap, where the Ninth Circuit held that design features (like a speed filter that encouraged reckless driving) are distinct from the third-party speech they carry, allowing plaintiffs to pursue claims based on negligent product design23. Thus, elevating AI curation to the status of protected speech successfully defeats government censorship laws but simultaneously destroys statutory immunity against private tort litigation31.

Part IV: Immunity vs. Regulation — The Doctrinal Battleground

The legal classification of AI outputs as protected speech, commercial speech, or non-expressive conduct carries profound structural consequences for the regulatory state. First Amendment doctrine acts as a formidable barrier to state action; if a regulation targets protected speech based on its content, it must survive "strict scrutiny," requiring the government to prove the law is narrowly tailored to achieve a compelling state interest using the least restrictive means available2.

The Risks of Classifying AI as Fully Protected Speech

If courts uniformly recognize generative AI outputs and algorithmic operations as fully protected First Amendment speech, regulators will face nearly insurmountable barriers6. This paradigm—often referred to by scholars as the "Lochnerization" of the First Amendment—would weaponize free speech doctrine to strike down consumer protection, antitrust, and technology regulations5.

1. "Slopigation" and the Chilling of Democratic Regulation: Generative AI has the capacity to produce synthetic content at scale without a human speaker. If this "AI slop" receives strict scrutiny protection, attempts by governments to regulate deceptive election deepfakes, algorithmic discrimination in housing and employment, or automated harassment would trigger massive First Amendment litigation3. Because the AI generates outputs far faster than human legal systems can process them, the sheer volume of protected machine speech could paralyze regulatory enforcement3.

2. Corporate Co-optation of the "Oppressed Speaker" Paradigm: The Supreme Court has recently shown profound solicitude for speakers resisting compelled expression, most notably in 303 Creative v. Elenis (2023)32. In 303 Creative, the Court ruled that Lorie Smith, a Christian web designer, could not be forced by Colorado's public accommodation laws to create custom websites for same-sex weddings that violated her religious beliefs32. Technology firms are already invoking 303 Creative to argue that algorithmic operations are equivalent to a designer's custom expression33. During oral arguments in NetChoice, industry lawyers explicitly cited 303 Creative to argue against state regulation of internet platforms35. If an AI developer is legally equated to the artisan in 303 Creative, the government cannot mandate that an AI company alter its weights to ensure unbiased or non-discriminatory outputs, as this would constitute unconstitutional compelled speech35. This stretches a doctrine designed to protect individual religious conscience to shield the algorithmic operations of trillion-dollar corporations33.

3. Defamation and Liability Avoidance: Applying traditional First Amendment defenses to AI outputs creates a legal absurdity. If an LLM hallucinates a defamatory statement about a public figure, traditional jurisprudence dictates that the plaintiff must prove "actual malice"—that the AI or its creator knew the statement was false or acted with reckless disregard for the truth4. Because an algorithm lacks consciousness and the developer cannot predict specific hallucinations arising from statistical probability distributions, proving actual malice is practically impossible, effectively immunizing AI companies from defamation liability2.

The Risks of Classifying AI as Non-Speech (Conduct)

Conversely, stripping First Amendment protections from AI outputs entirely introduces equally severe risks. If generative AI is classified strictly as non-expressive conduct or a mere software product, governments are empowered to regulate it under a highly deferential "rational basis" review6. Without constitutional protection, state governments could mandate that AI companies hard-code their models to suppress disfavored political views, scientific research, or controversial literature. Authoritarian or hyper-partisan legislatures could ban AI models that acknowledge climate change, provide reproductive health information, or criticize state policies. Furthermore, if the AI is denied protection, the human users who rely on generative AI as an expressive tool to create art, write code, or draft political essays are severely hindered9. Denying protection to the medium inherently chills the speech of the human relying on it9.

Intermediate Scrutiny: Commercial and Professional Speech Frameworks

To avoid the binary trap of strict scrutiny versus zero protection, courts and policymakers can adapt existing, lower-tier First Amendment frameworks to generative AI.

1. Commercial Speech and Compelled Disclosures

If an AI system is deployed by a corporation for profit, its outputs might be appropriately analyzed under the Central Hudson standard for commercial speech, which requires only "intermediate scrutiny"36. Under intermediate scrutiny, governments have broader latitude to regulate deceptive practices and ensure consumer safety38. Furthermore, regulations requiring AI companies to watermark synthetic content, label deepfakes, or disclose training data practices could be validated under the Zauderer doctrine39. Zauderer v. Office of Disciplinary Counsel (1985) established that the government can compel commercial speakers to disclose purely factual and uncontroversial information to prevent consumer deception40. Mandating that an AI platform disclose when content is synthetically generated fits neatly within the Zauderer framework, preserving the tool's utility while mitigating democratic harm and deceptive practices27.

2. Professional Speech Doctrine

When generative AI provides medical diagnoses or legal counsel, it encroaches upon regulated professional domains. In NIFLA v. Becerra (2018), the Supreme Court acknowledged that professional speech—recommendations and advice provided by licensed professionals to clients—is subject to distinct regulatory boundaries42. The practice of medicine, for instance, is largely an "informational service," heavily reliant on speech42. If an AI acts as a digital physician or attorney, its output ceases to be core political expression and becomes professional conduct43. Governments possess traditional police powers to regulate the unauthorized practice of law and medicine to protect public health and safety42. Therefore, state intervention to prohibit or strictly regulate AI systems from delivering unverified professional advice does not violate the First Amendment, as the state is regulating the conduct of the profession rather than the content of the expression42.

Part V: Synthesizing a Modern Framework for Algorithmic Expression

The Founders who drafted the First Amendment could not have conceived of a landscape where probabilistic algorithms generate persuasive, human-like language at an industrial scale, nor a society where autonomous agents curate the reality of billions of citizens1. Attempting to force generative AI into an eighteenth-century doctrinal framework designed for human moral agency produces legal paradoxes that either grant tech conglomerates unassailable immunity or hand the government a mandate for draconian censorship. A nuanced synthesis of constitutional doctrine must adapt to protect human users without outright banning the expressive tools they rely on, while simultaneously refusing to grant human constitutional rights to statistical software. This synthesis requires a tiered, human-centric threshold test based on the principles of traceability, autonomy, and design liability.

A Tiered Constitutional Framework

1\. The Human Causal Nexus (The Prompt and The Tool) The First Amendment must unequivocally protect the human user's input and the resulting immediate output when the AI acts as a direct expressive tool9. When a human deliberately prompts an AI to generate an essay, a piece of art, or computer code, the human is the speaker, and the AI is the medium10. Any government regulation seeking to ban human access to these tools, or to dictate the specific political viewpoints a model may generate in response to a human user's prompt, must trigger strict scrutiny. The tool is an extension of the human mind, and restricting it restricts human liberty. 2\. The Attenuation of Protection in Autonomous Systems First Amendment protection must inversely correlate with the autonomy of the system. As human direction fades, so does constitutional shelter22. When an AI system operates autonomously—such as an agentic system that independently initiates web campaigns, generates unprompted content, or relentlessly optimizes algorithmic feeds for engagement via dopamine loops without regard for human editorial standards—it crosses the threshold from expressive tool to behavioral product22. Autonomous "AI slop" and engagement-maximizing algorithms devoid of human editorial guidelines should be treated as non-expressive conduct, empowering democratic institutions to regulate them for safety, transparency, and consumer protection3. 3\. Product Liability and the Section 230 Realignment Following the inescapable logic of Anderson v. TikTok and Lemmon v. Snap, the legal system must differentiate between the content generated and the design architecture of the algorithm23. If technology companies successfully assert that algorithmic curation and generation are their own "first-party expression" protected by the First Amendment, they must inherently forfeit Section 230 immunity for the structural defects of that expression17. Courts must recognize a new frontier of design defect litigation, holding developers liable when their models are negligently designed to hallucinate defamatory falsehoods, bypass safety rails, or actively recommend inherently dangerous actions to minors23.

Conclusion

The First Amendment was designed to preserve the liberty of human conscience, foster individual moral agency, and ensure the vitality of democratic self-governance1. It was never intended to serve as an impenetrable corporate shield for the autonomous operations of probabilistic machines. To survive the era of generative AI and algorithmic curation, constitutional jurisprudence must firmly anchor speech rights to human intentionality. By treating AI as an expressive tool when directly controlled by humans, the law can protect the creative and intellectual liberties of users. Simultaneously, by classifying autonomous algorithmic generation and engagement optimization as commercial products subject to strict liability and democratic regulation, society can protect itself against the destabilizing forces of synthetic disinformation and behavioral manipulation. The future of free expression relies not on granting personhood to the machine, but on fiercely defending the agency, safety, and autonomy of the human holding the reins.

Works cited

1. The Declaration In the Age of General Purpose AI, https://law.stanford.edu/stanford-legal/the-declaration-at-250-the-declaration-in-the-age-of-general-purpose-ai/

2. An Introduction to the First Amendment (Chapter 13), https://www.cambridge.org/core/books/our-artificial-future/an-introduction-to-the-first-amendment/753D088DE12714CE527861677983A490

3. Of Slop and Swarms: The First Amendment's Next Test, https://knightcolumbia.org/blog/of-slop-and-swarms-the-first-amendments-next-test

4. Digital Doppelgängers: Rewriting Free Speech for the Age of AI, https://hulr.org/fall-2025-winners/digital-doppelgngers-rewriting-free-speech-for-the-age-of-ai

5. From Gods to Google \- Colorado Law Scholarly Commons, https://scholar.law.colorado.edu/context/faculty-articles/article/2669/viewcontent/2025\_Kaminski\_From\_Gods\_to\_Google.pdf

6. The First Amendment and GenAI Exceptionalism \- ResearchGate, https://www.researchgate.net/publication/412804941\_The\_First\_Amendment\_and\_GenAI\_Exceptionalism

7. Constructing AI Speech \- Colorado Law Scholarly Commons, https://scholar.law.colorado.edu/context/faculty-articles/article/2638/viewcontent/2024\_kaminski\_constructing\_ai\_speech.pdf

8. Does AI Have Speech Rights? \- Columbia Journalism Review, https://www.cjr.org/ask-anika/does-ai-have-speech-rights-journalists-using-ai-first-amendment-protections-character-openai-encourage-suicide-mass-shooting-cases.php

9. How does the First Amendment apply to AI? \- FIRE.org, https://www.fire.org/news/how-does-first-amendment-apply-ai

10. A Free Speech Dilemma: AI Outputs and the Constitution, https://ideaexchange.uakron.edu/cgi/viewcontent.cgi?article=2623\&context=akronlawreview

11. Freedom of Expression & Generative AI Outputs, https://digitalcommons.law.ou.edu/cgi/viewcontent.cgi?article=1679\&context=fac\_articles

12. Volume and Issue: Vol. 98, No. 5 \- Southern California Law Review, https://southerncalifornialawreview.com/volumeandissue/vol-98-no-5/

13. "AI Outputs are Not Protected Speech" with Professor Peter Salib, https://hls.harvard.edu/events/ai-outputs-are-not-protected-speech-with-professor-peter-salib-berkman-klein-center-x-aisst-ai-governance-speaker-series/

14. arXiv:2405.20362v1 \[cs.CL\] 30 May 2024, https://arxiv.org/pdf/2405.20362

15. A Constitutional Black Box \- SCARAB Bates, https://scarab.bates.edu/cgi/viewcontent.cgi?article=1022\&context=bulr

16. Moody v. NetChoice: The Supreme Court Meets the Free Speech, https://www.journals.uchicago.edu/doi/10.1086/735007

17. The Third Circuit's Section 230 Decision In Anderson v. TikTok Is, https://www.techdirt.com/2024/09/03/the-third-circuits-section-230-decision-in-anderson-v-tiktok-is-pure-poppycock/

18. MOODY v. NETCHOICE, LLC | Supreme Court \- Law.Cornell.Edu, https://www.law.cornell.edu/supremecourt/text/22-277

19. Moody v. NetChoice, LLC | 603 U.S. \_\_\_ (2024) \- Justia Law, https://supreme.justia.com/cases/federal/us/603/22-277/

20. Is Internet Content Too Engaging? \- Lawfare, https://www.lawfaremedia.org/article/is-internet-content-too-engaging

21. Judges rule Big Tech's free ride on Section 230 is over | Hacker News, https://news.ycombinator.com/item?id=41391868

22. Is an algorithmic feed 'speech'? Courts are about to decide, https://www.dailyjournal.com/articles/393748-is-an-algorithmic-feed-speech-courts-are-about-to-decide

23. Beyond the warning label: Why 'design defect' is the new frontier in, https://www.dailyjournal.com/mcle/1875-beyond-the-warning-label-why-design-defect-is-the-new-frontier-in-social-media-litigation

24. Four Key Takeaways from the Moody v. NetChoice and ... \- Epic.org, https://epic.org/four-key-takeaways-from-the-netchoice-v-moody-and-paxton-oral-arguments/

25. 1 Prepared remarks of James Fallows Tierney before the meeting of, https://www.sec.gov/files/spotlight/iac/tierney-remarks-iac062223pdf.pdf

26. ANDERSON v. TIKTOK INC (2024) \- FindLaw Caselaw, https://caselaw.findlaw.com/court/us-3rd-circuit/116519493.html

27. Beyond Section 230: Principles for AI Governance, https://harvardlawreview.org/print/vol-138/beyond-section-230-principles-for-ai-governance/

28. Anderson v. TikTok: a New Challenge to § 230 Immunity, https://sjipl.mainelaw.edu/2024/11/01/anderson-v-tiktok-a-new-challenge-to-%C2%A7-230-immunity/

29. Lawsuit Against Snapchat Rightfully Goes Forward Based on, https://www.eff.org/deeplinks/2021/05/lawsuit-against-snapchat-rightfully-goes-forward-based-speed-filter-not-user

30. How Recent Court Rulings Have Torched Big Tech's Litigation, https://lawreview.unl.edu/smoke-how-recent-court-rulings-have-torched-big-techs-litigation-strategy-and-why-it-matters/

31. Between a TikTok and a Hard Place: Products Liability, Section 230, https://truthonthemarket.com/2024/09/12/between-a-tiktok-and-a-hard-place-products-liability-section-230-and-the-first-amendment/

32. 303 Creative LLC v. Elenis \- Hamilton Lincoln Law Institute, https://hlli.org/303-creative-llc-v-elenis/

33. How Religious Speech Cases May Fortify the Deregulatory First, https://digitalcommons.du.edu/cgi/viewcontent.cgi?article=1696\&context=law\_facpub

34. Immoral minority: Why SCOTUS puts First Amendment rights for, http://www.milwaukeeindependent.com/syndicated/immoral-minority-scotus-puts-first-amendment-rights-christians-equal-protection/

35. Who Decides What AI Decides \- Loyola Marymount University, https://www.lls.edu/faculty/lls-impact/whodecideswhataidecides/

36. Is freedom of expression a First Amendment right? \- Michael, https://michaelcarbonara.com/is-freedom-of-expression-a-first-amendment-right/

37. Court denies xAI's bid to block California AI training data law, https://ppc.land/court-denies-xais-bid-to-block-california-ai-training-data-law/

38. California Chasing Arrows Recycling Law, SB 343, Preliminarily, https://quickreads.ext.katten.com/post/102nc64/california-chasing-arrows-recycling-law-sb-343-preliminarily-enjoined

39. California AI Model Training Disclosure Law Likely Doesn't Violate, https://reason.com/volokh/2026/03/10/california-ai-model-training-disclosure-law-likely-doesnt-violate-first-amendment/

40. Required Deepfake Labeling: Contradicting or Supporting The First, https://www.culawreview.org/journal/required-deepfake-labeling-contradicting-or-supporting-the-first-amendment

41. In Too Deep: Navigating an Unpredictable Algorithm, https://digitalcommons.tourolaw.edu/cgi/viewcontent.cgi?article=3511\&context=lawreview

42. What Is Private Law? (Part I) \- Health Law as Private Law, https://www.cambridge.org/core/books/health-law-as-private-law/what-is-private-law/44AB3F1D9CA4E909E22A643F31BF218A

43. Free Speech or Professional Conduct? The Supreme Court, https://heinonline.com/blog/2025/10/free-speech-or-professional-conduct-the-supreme-court-examines-colorados-ban-on-conversion-therapy-for-minors/