First Amendment & AI
Lecture 5 • SPI/COS 352

First Amendment

Peter Henderson, J.D., Ph.D.
Assistant Professor of Computer Science and Public Affairs

Slides in part based on materials from

Credit image 3

Disclaimer

Please be respectful when discussing scenarios. I will avoid potentially offensive content when possible, but free speech doctrine necessarily covers some potentially offensive speech. Please feel free to step out as needed.

What is the text of the First Amendment?

"Congress shall make no law respecting an establishment of religion, or prohibiting the free exercise thereof; or abridging the freedom of speech, or of the press; or the right of the people peaceably to assemble, and to petition the Government for a redress of grievances."

RELIGION CLAUSES
SPEECH & PRESS
ASSEMBLY & PETITION

Today we will mainly focus on speech.

"Congress shall make no law... abridging the freedom of speech, or of the press

SPEECH & PRESS

But see great article on interaction between religion cases and tech cases

From Gods to Google article

"From Gods to Google" - Yale Law Journal

The Right Only Applies to the Government

The First Amendment largely does not protect speakers against private individuals or organizations, such as private employers, or private landowners. The First Amendment restrains only the government.

Image from page 8

The Right is Broad, but not Unlimited

But the First Amendment largely does not allow the government to regulate the content of anyone's speech with exceptions.

Image from page 8

First Amendment Speech Protection Analysis

Protected Speech
Content-Based Restriction
IF restriction targets subject matter or viewpoint
THEN Strict Scrutiny Required
  • Compelling government interest
  • Narrowly tailored regulation
  • Least restrictive means used
Reed v. Town of Gilbert, 576 U.S. 155, 163 (2015) (content-based laws are "presumptively unconstitutional"); Sable Communications of Cal. v. FCC, 492 U.S. 115, 126 (1989) (least restrictive means requirement)
Content-Neutral Restriction
IF restriction regulates time, place, & manner
THEN Forum Analysis Required
Public Forum
IF Public Forum (e.g., streets, sidewalks, public parks) or Designated Public Forum (e.g., ads on city buses)
THEN Intermediate Scrutiny
  • Must be subject AND viewpoint neutral
  • Narrowly tailored, important interest
  • Leaves open alternative channels
Limited/Non-Public Forum
IF Limited Public Forum or Non-Public Forum (e.g., military base, airports)
THEN Rational Basis
  • Must be viewpoint neutral
  • Reasonably related to legitimate purpose
Unprotected Speech
Unprotected Speech
IF speech advocates violence, fighting words, hostile audience speech, obscene speech, or defamatory speech
THEN No First Amendment Protection
Chaplinsky v. New Hampshire, 315 U.S. 568, 571-72 (1942) (certain well-defined categories of speech not essential to ideas); Brandenburg v. Ohio, 395 U.S. 444 (1969) (incitement test)
Lesser Protected Speech - Government/Public Employees
Public Employee Speech
IF employee speaking as private citizen on matter of public concern
THEN Pickering Balancing Test
  • Balance employee rights vs. government efficiency
  • Adequate justification for different treatment
Not Public Concern
IF not of public concern
THEN Not Protected (wide deference to employer)
Official Duties
IF statements made pursuant to official duties
THEN Not Protected
Pickering v. Bd. of Educ., 391 U.S. 563, 568 (1968); Connick v. Myers, 461 U.S. 138, 147 (1983); Garcetti v. Ceballos, 547 U.S. 410, 421 (2006)
Lesser Protected Speech - Commercial Speech
Commercial Speech
IF speech is commercial (advertising/commercial communications)
THEN Central Hudson Test (Intermediate Scrutiny)
  • Speech must concern lawful activity and not be misleading
  • Government interest must be substantial
  • Regulation must directly advance that interest
  • Regulation must be no more extensive than necessary
Central Hudson Gas & Elec. Corp. v. Pub. Serv. Comm'n, 447 U.S. 557, 563-66 (1980) (establishing four-part test)

Facial Challenge vs. "As Applied" Challenge

Facial Challenge

"a substantial number of [the law's] applications are unconstitutional, judged in relation to the statute's plainly legitimate sweep."

Americans for Prosperity Foundation v. Bonta, 594 U. S. 595, 615

"As Applied" Challenge

Unconstitutional as applied to these (or similar) facts.

Schenck v. United States (1919)

Two socialists, Charles Schenck and Elizabeth Baer, distributed roughly 15,000 leaflets to army conscripts during World War I that argued that conscription violated the 13th Amendment and urged them to assert their rights.

The two were charged with, and ultimately convicted of, impeding military recruitment, in violation of the Espionage Act that Congress passed shortly after the United States entered World War I.

Image from page 13

Schenck v. United States (1919)

Schenck and Baer challenged their convictions on First Amendment grounds, but the Supreme Court—in an opinion written by Holmes—held that, because the pamphlets could have plausibly impeded the war effort, the two defendants could be punished for their speech.

Image from page 14

Schenck v. United States (1919)

Famous "Fire in a crowded theater" meme

Image from page 15

BRANDENBURG V. OHIO (1969)

Brandenburg, a leader in the Ku Klux Klan, made a speech at a Klan rally and was later convicted under an Ohio criminal syndicalism law. The law made illegal advocating "crime, sabotage, violence, or unlawful methods of terrorism as a means of accomplishing industrial or political reform," as well as assembling "with any society, group, or assemblage of persons formed to teach or advocate the doctrines of criminal syndicalism."

Did Ohio's criminal syndicalism law, prohibiting public speech that advocates various illegal activities, violate Brandenburg's right to free speech as protected by the First and Fourteenth Amendments?

Image from page 16

BRANDENBURG V. OHIO (1969)

Brandenburg Test: Speech may be prohibited if it is directed to incite and likely to incite imminent lawless action.

Overruled Schenck!

The criminal syndicalism act made illegal the advocacy and teaching of doctrines while ignoring whether or not that advocacy and teaching would actually incite imminent lawless action. The failure to make this distinction rendered the law overly broad and in violation of the Constitution.

Image from page 17

New York Times v. Sullivan

1960: New York Times ran full-page advertisement paid for by civil rights activists

Ad criticized Montgomery, Alabama police force for treatment of civil rights activists

Supposedly included some inaccurate information regarding the civil rights protesters and the Alabama police—stating that the police brutally assaulted them

L. B. Sullivan (police commissioner in Montgomery) took offense and sued the New York Times for libel

Image from page 18

New York Times v. Sullivan

March 9, 1964: Court ruled in favor of New York Times

Court ruled that the Times would have to have had "actual malice"

"With knowledge that [information] was false or with reckless disregard for the truth"

Sullivan could only recover damages if he proved libelous statement had a distinct goal of defaming his character

Revolutionary impact on 1st Amendment:

Press now has "right to be wrong" if not acting recklessly or maliciously

Some completely false statements (if not deliberate) could be entirely free from punishment (new establishment under New York Times v. Sullivan)

Image from page 19

Lamont v. Postmaster General (1965)

"Mail matter, except sealed letters, which originates or which is printed or otherwise prepared in a foreign country and which is determined by the Secretary of the Treasury pursuant to rules and regulations to be promulgated by him to be 'communist political propaganda', shall be detained by the Postmaster General upon its arrival for delivery in the United States, or upon its subsequent deposit in the United States domestic mails, and the addressee shall be notified that such matter has been received and will be delivered only upon the addressee's request, except that such detention shall not be required in the case of any matter which is furnished pursuant to subscription or which is otherwise ascertained by the Postmaster General to be desired by the addressee." 39 U.S.C. 4008 (1964)

Image from page 20

Lamont v. Postmaster General (1965)

Dr. Corliss Lamont (left), founder and first chairman of Council on American-Soviet Friendship in the 1940s, had a copy of the Peking Review detained and declined to respond to the government's inquiry as to whether he wished to receive the delivery.

Lamont subsequently filed suit alleging that Section 4008 violated his 1st Amendment and 5th Amendment rights.

Image from page 21

Lamont v. Postmaster General (1965)

Justice Brennan, "It would be a barren marketplace of ideas that had only sellers and no buyers."

The Act, as construed and applied, is unconstitutional, since it imposes on the addressee an affirmative obligation which amounts to an unconstitutional limitation of his rights under the First Amendment.

Image from page 22

Kleindienst v. Mandel (1972)

Facts: Ernest Mandel, a Belgian journalist and Marxist theoretician, was invited by American scholars to participate in academic conferences in the United States.

Under the Immigration and Nationality Act of 1952, Mandel was deemed ineligible for entry due to his advocacy of communist doctrines.

Attorney General Kleindienst declined to waive this ineligibility, citing Mandel's unscheduled activities during a previous U.S. visit when a waiver had been granted.

Question: Did Attorney General Kleindienst violate the First Amendment rights of the scholars and students who invited Mandel to the United States by refusing to allow Mandel to enter the country?

Discuss! (3 Min)

Image from page 23

Kleindienst v. Mandel (1972)

No. In a 6-3 decision written by Justice Harry A. Blackmun, the Court held that the First Amendment did not obligate Kleindienst to grant a waiver to Mandel. Justice Blackmun acknowledged that the First Amendment protected the scholars and students' right to listen to and engage with Mandel in person --and that Mandel's participation via telephone was not a good replacement-- but held that the implication of First Amendment rights was not dispositive here.

Justice Blackmun noted Congress' longstanding power to exclude aliens from the United States, and to set the terms and conditions of their entry. Through the Immigration and Nationality Act, Congress legitimately delegated to the executive the authority to waive a finding of inadmissibility. He described the historical pattern of increasing federal control on the admissibility of aliens, particularly regarding individuals with Communist affiliation or views. Justice Blackmun held that the Court would not intervene so long as the executive used its waiver power on the basis of a facially legitimate and bona fide reason. This test did not balance the First Amendment interests of persons seeking to communicate with the applicant.

Image from page 24

Compare Kleindienst against Lamont. Where do you draw the line?

(Discuss 3 minutes.)

Panel: What does Peter Salib say about these cases?

Peter Salib draws a connection between speakers outside of the United States and AI. Do you agree? Do you think his reading of Kleindienst is correct? What are the risks to non-AI speech from adopting this viewpoint?

(Discuss 3 minutes.)

Break

Moody v. Netchoice (2024)

In 2021, Florida and Texas enacted statutes regulating large social-media companies and other internet platforms. The States' laws differ in the entities they cover and the activities they limit.

But both curtail the platforms' capacity to engage in content moderation—to filter, prioritize, and label the varied third-party messages, videos, and other content their users wish to post.

Both laws also include individualized explanation provisions, requiring a platform to give reasons to a user if it removes or alters her posts

https://knightcolumbia.org/cases/netchoice

Image from page 30

Moody v. Netchoice (2024) - Texas Statute

Texas Statute: With several exceptions, the statute prevents platforms from "censor[ing]" a user or a user's expression based on viewpoint.

That ban on "censor[ing]" covers any action to "block, ban, remove, deplatform, demonetize, deboost, restrict, deny equal access or visibility to, or otherwise discriminate against expression."

The statute also requires that "concurrently with the removal" of user content, the platform shall "notify the user" and "explain the reason the content was removed."

The user gets a right of appeal, and the platform must address an appeal within 14 days.

https://knightcolumbia.org/cases/netchoice

Image from page 31

Moody v. Netchoice (2024) - Florida Statute

Florida Statute: The statute restricts varied ways of "censor[ing]" or otherwise disfavoring posts—including deleting, altering, labeling, or deprioritizing them—based on their content or source. §501.2041(1)(b).

For example, the law prohibits a platform from taking those actions against "a journalistic enterprise based on the content of its publication or broadcast." §501.2041(2)(j).

Similarly, the law prevents deprioritizing posts by or about political candidates. See §501.2041(2)(h). And the law requires platforms to apply their content-moderation practices to users "in a consistent manner." §501.2041(2)(b).

https://knightcolumbia.org/cases/netchoice

Image from page 33
Image from page 34
Image from page 35
Image from page 36 Image from page 36
Image from page 37

Do you agree or disagree with SCOTUS? Could Texas or Florida modify their laws to be compatible with the First Amendment?

(Discuss 3 minutes.)

Lochner.com article page

Justice Amy Coney Barrett Concurrence

"What if a platform's owners hand the reins to an AI tool and ask it simply to remove 'hateful' content? If the AI relies on large language models to determine what is 'hateful' and should be removed, has a human being with First Amendment rights made an inherently expressive 'choice ... not to propound a particular point of view'?"

"In other words, technology may attenuate the connection between content-moderation actions (e.g., removing posts) and human beings' constitutionally protected right to 'decide for [themselves] the ideas and beliefs deserving of expression, consideration, and adherence.'"

Justices Alito & Gorsuch Concurring Opinions

"[W]hen AI algorithms make a decision, 'even the researchers and programmers creating them don't really understand why the models they have built make the decisions they make.'"

"Are such decisions equally expressive as the decisions made by humans? Should we at least think about this?"

NetChoice v. Bonta

Facts: California's Age-Appropriate Design Code Act (CAADCA) requires online platforms to assess and mitigate risks to children, prohibit use of minors' personal information in detrimental ways, and configure default privacy settings for high privacy.

NetChoice, representing tech companies, challenged the CAADCA as violating the First Amendment through content-based regulation of online speech and imposing prior restraints on expression.

Holding: The Ninth Circuit upheld parts of a preliminary injunction (August 2024), finding that certain CAADCA provisions likely violate the First Amendment by imposing content-based restrictions that may not survive strict scrutiny.

Discuss for 3 minutes, agree or disagree? Why? Why not?

Image from page 40

Garcia v. Character Technologies, Inc.

Facts: Megan Garcia sued Character Technologies, Inc. alleging that her 14-year-old son, Sewell Setzer III, developed an emotional attachment to a Character.AI chatbot named "Dany" that led to his suicide.

First Amendment Argument: Character.AI argued that AI chatbot outputs are protected speech under the First Amendment, and restricting them would infringe on free speech rights.

Court's Decision: U.S. District Judge Anne Conway denied Character.AI's motion to dismiss (May 2025), stating the court is "not prepared to hold that Character.AI's output is speech" at this stage.

Image from page 41

Consider the case law and 1A principles you've seen so far, discuss with your neighbors coming up with arguments for and against. Also discuss design decisions that might sway your analysis, then we'll go around the room and think about how designs might change the speech analysis.

Jawboning

Government officials using persuasion or pressure to influence private entities' actions without formal legislation or regulation.

In the context of social media, this involves government pressure on platforms to moderate content in ways that align with government preferences.

Murthy v. Missouri (2024)

Facts: Missouri and Louisiana attorneys general sued federal officials, including Surgeon General Vivek Murthy, alleging government coercion of social media platforms.

The states claimed federal officials pressured platforms to suppress content about COVID-19 vaccines, election integrity, and other politically sensitive topics.

Individual social media users joined as plaintiffs, claiming their posts were removed due to government pressure on platforms.

Question: Did federal officials violate the First Amendment by coercing social media platforms to suppress certain viewpoints?

Discuss! (3 Min)

Lakier article

Murthy v. Missouri (2024) - Holding

Holding: The Court held that the plaintiffs lacked standing to sue because they failed to demonstrate a direct causal link between the government's actions and the platforms' content moderation decisions.

To establish standing, plaintiffs must show a substantial risk that, in the near future, at least one platform will restrict the speech of at least one plaintiff in response to the actions of at least one government defendant.

The Court found this to be a "tall order" on the record presented.

The Court emphasized that plaintiffs did not provide sufficient evidence to show that their past injuries were traceable to the government's conduct, nor did they establish a substantial risk of future harm attributable to the government.

Lakier article

National Rifle Association v. Vullo (2024)

Key Facts (as alleged): After a whistleblower said the NRA's "Carry Guard" insurance violated NY law, NY's DFS Superintendent Maria Vullo investigated three insurers linked to the product. Following the Parkland shooting, she pressed those insurers to end all NRA-"affinity" business.

In a private meeting she allegedly offered leniency on unrelated "technical regulatory infractions" if an insurer cut ties with the NRA; publicly, she issued Guidance Letters praising "social backlash" against the NRA, warning about "reputational" and other risks of doing business with it.

Holding: A unanimous Court reaffirmed Bantam Books: government "cannot do indirectly what [it] is barred from doing directly." Officials violate the First Amendment when they intentionally use threats of regulatory harm (or promises of benefits) to pressure private companies to cut ties with disfavored speakers/associations.

Implications: It rejects lower-court tests requiring near-compulsion. Any scheme of constitutional evasion via threats/promises is out of bounds, regardless of degree or result. Expect a shift toward an intent-focused, categorical anti-jawboning rule.

Lakier article

Implications of Murthy v. Missouri

1. Dismissal on Standing and Causation Requirement:

The Murthy case (7-2 decision) was dismissed on standing, notably "without engaging Bantam or Vullo." For speakers who claim their content was suppressed due to coerced platforms, the Court now requires "evidence of specific causation." This requirement "raises the bar at the jurisdictional (Article III) stage."

2. Scope of Causation Requirement:

The causation requirement "does not apply across the board." If the "direct target" of the "jawboning" (e.g., a business or platform) sues, the injury is considered the threatened interference with "editorial autonomy." In such cases, success is not required to be unconstitutional, making standing "easier" for these plaintiffs.

3. Redressability for Injunctive Relief:

For speakers seeking "injunctive relief," Murthy also "tightens redressability." Plaintiffs must demonstrate "ongoing" government pressure or prove that the platform had "no independent reasons" to moderate their content. This is "often hard to prove given mixed motives and shifting policies."

Lakier article

Trump Executive Order on "Woke AI"

Trump Executive Order on Woke AI

Crossroads: Locherization vs. Government Intervention

Forked Road Meme