Amicus Curiae Brief — Twitter, Inc., Petitioner v. Mehier Taamneh, et al.
Supreme Court briefDec 5, 2022
Ask Donna
What actually matters in this document.
Text
No. 21-1496
IN THE
Supreme Court of the United States
————
TWITTER, INC.,
Petitioner,
v.
MEHIER TAAMNEH; LAWRENCE TAAMNEH;
SARA TAAMNEH; DIMANA TAAMNEH,
Respondents.
————
On Writ of Certiorari to the
United States Court of Appeals
for the Ninth Circuit
————
BRIEF OF CENTER FOR DEMOCRACY &
TECHNOLOGY, AMERICAN CIVIL LIBERTIES
UNION, AMERICAN CIVIL LIBERTIES UNION
FOUNDATION OF NORTHERN CALIFORNIA,
ELECTRONIC FRONTIER FOUNDATION,
KNIGHT FIRST AMENDMENT INSTITUTE AT
COLUMBIA UNIVERSITY, R STREET
INSTITUTE, AND REPORTERS COMMITTEE
FOR FREEDOM OF THE PRESS
AS AMICI CURIAE
IN SUPPORT OF PETITIONER
————
Samir Jain
Emma Llansó
Caitlin Vogus
CENTER FOR DEMOCRACY &
TECHNOLOGY
1401 K Street NW, Suite 200
Washington, DC 20005
David D. Cole
AMERICAN CIVIL LIBERTIES
UNION FOUNDATION
915 15th Street NW
Washington, DC 20005
Kathleen R. Hartnett
Counsel of Record
Benjamin Kleine
Reece Trevor
COOLEY LLP
3 Embarcadero Center
20th Floor
San Francisco, CA 94111
(415) 693-2000
khartnett@cooley.com
Patrick J. Hayden
COOLEY LLP
55 Hudson Yards
New York, NY 10001
Counsel for Amici Curiae
[Additional Counsel Listed On Inside Cover]
WILSON-EPES PRINTING CO., INC. – (202) 789-0096 – WASHINGTON, D.C. 20002
Jennifer Stisa Granick
AMERICAN CIVIL LIBERTIES
UNION FOUNDATION
39 Drumm Street
San Francisco, CA 94111
Aaron Mackey
ELECTRONIC FRONTIER
FOUNDATION
815 Eddy Street
San Francisco, CA 94109
Patrick Toomey
Sarah Taitz
Elizabeth Gyori
AMERICAN CIVIL LIBERTIES
UNION FOUNDATION
125 Broad Street
New York, NY 10004
Bruce D. Brown
Katie Townsend
Gabe Rottman
Grayson Clary
REPORTERS COMMITTEE FOR
FREEDOM OF THE PRESS
1156 15th Street NW,
Suite 1020
Washington, DC 20005
Nicole A. Ozer
AMERICAN CIVIL LIBERTIES
UNION FOUNDATION OF
NORTHERN CALIFORNIA
39 Drumm Street
San Francisco, CA 94111
Chris Riley
Jonathan Cannon
THE R STREET INSTITUTE
1212 New York Avenue NW,
Suite 900
Washington, DC 20005
Alex Abdo
Anna Diakun
Alexia Ramirez
KNIGHT FIRST AMENDMENT
INSTITUTE AT COLUMBIA
UNIVERSITY
475 Riverside Drive,
Suite 302
New York, NY 10115
i
TABLE OF CONTENTS
Page
INTERESTS OF AMICI CURIAE ............................. 1
INTRODUCTION AND SUMMARY OF
ARGUMENT .................................................... 3
ARGUMENT ............................................................... 7
I.
II.
Interpretation of the Anti-Terrorism Act
Must Account for the Statute’s Effects
on Speech ......................................................... 7
A.
The First Amendment Requires
Robust Protection for
Intermediaries of Speech ...................... 7
B.
Where Aiding-and-Abetting
Liability Under the AntiTerrorism Act Is Premised on
Speech, the Act Should Be
Interpreted to Avoid Chilling the
Protected Speech of Online
Platforms and Their Users ................. 15
The Ninth Circuit’s Overly Broad
Interpretation of ATA Liability as
Applied to Speech Will Result in the
Suppression of Lawful Speech....................... 18
A.
The Ninth Circuit’s Construction
Incentivizes Intermediaries to
Over-Moderate Content, Causing
Dramatic Suppression of
Protected Speech ................................. 19
ii
TABLE OF CONTENTS
(continued)
Page
B.
The Ninth Circuit’s Construction
Incentivizes Intermediaries to
Impose Blanket Bans on
Controversial Topics Like
Terrorism, Barring Speech
Regardless of Its Constitutional
Protection ............................................ 23
CONCLUSION ......................................................... 25
iii
TABLE OF AUTHORITIES
Page(s)
Cases
Backpage.com, LLC v. McKenna,
881 F. Supp. 2d 1262 (W.D. Wash.
2012) ............................................................... 13, 14
Bantam Books, Inc. v. Sullivan,
372 U.S. 58 (1963) .................................... 10, 11, 23
Brandenburg v. Ohio,
395 U.S. 444 (1969) ................................................ 7
City of Houston v. Hill,
482 U.S. 451 (1987) .............................................. 12
Cohen v. California,
403 U.S. 15 (1971) ................................................ 12
Cubby v. CompuServe,
776 F. Supp. 135 (S.D.N.Y. 1991) ........................ 13
Ginsberg v. New York,
390 U.S. 629 (1968) .............................................. 11
Gonzalez v. Google LLC,
2 F.4th 871 (9th Cir. 2021) .................................. 19
Halberstam v. Welch,
705 F.2d 472 (D.C. Cir. 1983) ................ 3, 4, 15, 16
Hamling v. United States,
418 U.S. 87 (1974) ................................................ 11
iv
TABLE OF AUTHORITIES
(continued)
Page(s)
Manual Enterprises, Inc. v. Day,
370 U.S. 478 (1962) .......................................... 9, 10
McCullen v. Coakley,
573 U.S. 464 (2014) .............................................. 12
New York Times Co. v. Sullivan,
376 U.S. 254 (1964) ................................................ 8
Rice v. Paladin Enterprises, Inc.,
128 F.3d 233 (4th Cir. 1997) ................................ 17
Scales v. United States,
367 U.S. 203 (1961) .............................................. 12
Smith v. California,
361 U.S. 147 (1959) ...................................... passim
Universal Comm’cn Sys., Inc. v. Lycos,
478 F.3d 413 (1st Cir. 2007)................................. 14
Statutes
18 U.S.C. § 2333 .......................................................... 4
47 U.S.C. § 230 .................................................... 13, 14
Allow States and Victims to Fight
Online Sex Trafficking Act/Stop
Enabling Sex Traffickers Act, Pub. L.
No. 115-164, 132 Stat. 1253 (2018) ..................... 23
v
TABLE OF AUTHORITIES
(continued)
Page(s)
Other Authorities
Abigail Moss, “Such a Backwards Step”:
Instagram Is Now Censoring Sex
Education Accounts, Vice (Jan. 8,
2021), https://tinyurl.com/5a4mezp9 ................... 23
About, Global Internet Forum to
Counter Terrorism,
https://gifct.org/about/ (last visited
Dec. 3, 2022) ......................................................... 18
Carey Shenkman et al., Do You See
What I See? Capabilities and Limits
of Automated Multimedia Content
Analysis, Center for Democracy &
Technology (May 2021),
https://tinyurl.com/yzw2hmah ............................. 22
Hannah Bloch-Wehba, Automation in
Moderation, 53 Cornell Int’l L.J. 41
(2020) .............................................................. 20, 21
Human Rights Watch (@hrw), Twitter
(Nov. 26, 2022),
https://tinyurl.com/ywa977c................................. 24
Interview with Osama Bin Laden, PBS
Frontline (May 1998),
https://perma.cc/9V9Y-UY9T ............................... 18
vi
TABLE OF AUTHORITIES
(continued)
Page(s)
Jack Shepherd, 22 Essential Twitter
Statistics You Need to Know in 2022,
Social Shepherd (Nov. 5, 2022),
https://tinyurl.com/5bjy4jw7 ................................ 20
Jon Porter, Instagram blames
“enforcement error” for removal of
posts about Al-Aqsa Mosque, The
Verge (May 13, 2021),
https://tinyurl.com/yn8kujej ................................ 22
Marvin Ammori, The “New” York Times:
Free Speech Lawyering in the Age of
Google and Twitter, 127 Harv. L.
Rev. 2259 (2014) ..................................................... 4
Mikael Thalen, YouTube is cracking
down on independent journalists who
covered the Capitol riot, Daily Dot
(Feb. 3, 2021),
https://tinyurl.com/3kdwbuhx.............................. 22
Nafia Chowdhury, Automated Content
Moderation: A Primer, Stanford
Freeman Spogli Institute (March 19,
2022), https://tinyurl.com/3txpeufd ..................... 21
Olivia Solon, “Facebook doesn’t care”:
Activists say accounts removed
despite Zuckerberg's free-speech
stance, NBC News (June 15, 2020),
https://tinyurl.com/4ve5nypm .............................. 22
vii
TABLE OF AUTHORITIES
(continued)
Page(s)
Pope Francis (@pontifex), Twitter (Sept.
15, 2022),
https://tinyurl.com/3ycm2hwd ............................. 24
Provisional I.R.A. Continues Truce, N.Y.
Times, Jan. 3, 1975,
https://tinyurl.com/vkat5zsb ................................ 18
Seth F. Kreimer, Censorship by Proxy:
The First Amendment, Internet
Intermediaries, and the Problem of
the Weakest Link, 155 U. Pa. L. Rev.
11 (2006) ............................................................... 12
Shannon Liao, Tumblr will ban all adult
content on December 17th, The
Verge (Dec. 3, 2018),
https://bit.ly/2SmoC5A ......................................... 24
Spandana Singh, Everything in
Moderation: An Analysis of How
Internet Platforms Are Using
Artificial Intelligence to Moderate
User Generated Content, New
America 5, 7 (July 15, 2019),
available at
https://tinyurl.com/23zhxmaa .............................. 21
INTERESTS OF AMICI CURIAE1
Amici curiae are organizations that support and
advocate for Internet users’ free expression and other
human rights. Amici have a strong interest in
ensuring that individuals are able to access and
participate in all forms of constitutionally protected
speech online, as the First Amendment guarantees.
The Center for Democracy & Technology (“CDT”)
is a non-profit public interest organization. For more
than 25 years, CDT has represented the public’s
interest in an open, decentralized Internet and worked
to ensure that the constitutional and democratic
values of free expression and privacy are protected in
the digital age. CDT regularly advocates before
legislatures, regulatory agencies, and courts in
support of First Amendment rights on the Internet
and other protections for online speech, including
limits on intermediary liability for user-generated
content.
The American Civil Liberties Union (“ACLU”) is
a nationwide, nonprofit, nonpartisan organization
dedicated to defending the principles embodied in the
Federal Constitution and our nation’s civil rights
laws. The ACLU of Northern California is the
Northern California affiliate of the ACLU. Since its
founding in 1920, the ACLU has frequently appeared
before this Court, the lower federal courts, and state
courts in cases defending Americans’ free speech and
1
No counsel for a party authored this brief in whole or in part,
and no person other than amici or their counsel made a monetary
contribution to this brief’s preparation and submission. All
parties have provided blanket consent to this filing.
2
freedom of association, including their exercise of
those rights online.
The Electronic Frontier Foundation (“EFF”) is a
member-supported,
nonprofit
civil
liberties
organization that has worked for more than thirty
years to protect innovation, free expression, and civil
liberties in the digital world. On behalf of its more
than 38,000 dues-paying members, EFF ensures that
users’ interests are presented to courts considering
crucial online free speech issues, including users’
rights to transmit and receive information online.
The Knight First Amendment Institute at
Columbia University is a non-partisan, not-for-profit
organization that works to defend the freedoms of
speech and the press in the digital age through
strategic litigation, research, and public education.
The Knight Institute’s aim is to promote a system of
free expression that is open and inclusive, that
broadens and elevates public discourse, and that
fosters creativity, accountability, and effective selfgovernment. Protecting the integrity and vitality of
online platforms as forums for public discourse is of
special concern to the Knight Institute.
The R Street Institute is a nonprofit, nonpartisan
public-policy research organization.
R Street’s
mission is to engage in policy research and educational
outreach that promotes free markets as well as limited
yet effective government, including properly
calibrated legal and regulatory frameworks that
support economic growth and individual liberty.
The Reporters Committee for Freedom of the
Press (“RCFP”) is an unincorporated nonprofit
association founded by leading journalists and media
lawyers in 1970 when the nation’s news media faced
3
an unprecedented wave of government subpoenas
forcing reporters to name confidential sources. Today,
RCFP provides pro bono legal representation, amicus
curiae support, and other legal resources to protect
First Amendment freedoms and the newsgathering
rights of journalists.
INTRODUCTION AND
SUMMARY OF ARGUMENT
This Court has long recognized that the First
Amendment forbids not just direct regulation of
protected speech, but also state action likely to result
in self-censorship and the curtailment of lawful
speech. The Ninth Circuit held that under the AntiTerrorism Act (“ATA”), an online platform with only
generalized awareness that a terrorist organization–
among literally hundreds of millions of other
speakers–may have used its service has “knowingly
provided substantial assistance” to an act of terrorism.
That interpretation threatens to substantially narrow
the speech that platforms host, raising serious First
Amendment concerns. But those concerns are easily
avoided by rejecting the Ninth Circuit’s erroneous
reading of the statute, and by requiring that a
defendant possess actual knowledge that a specific
piece of user-generated content on its platform
provides substantial assistance to a terrorist act
before imposing aiding-and-abetting liability on the
basis of its function as a speech intermediary.
The multi-factor Halberstam v. Welch test that
Congress specified for determining aiding-andabetting liability under the ATA can and should be
interpreted to account for the First Amendment
implications of imposing liability on speech
“intermediaries”–third parties who publish or
distribute the speech of others–like Petitioner
4
Twitter, Inc. and Respondents Facebook, Inc. and
These platforms and other
Google LLC.2
intermediaries provide essential fora for speech and
have become a primary source of news, information,
and discussions across the nation and around the
world. Indeed, “[t]he numbers suggest that companies
like Google and Twitter have at least as great an
impact on free expression as do traditional
newspapers.” Marvin Ammori, The “New” York
Times: Free Speech Lawyering in the Age of Google
and Twitter, 127 Harv. L. Rev. 2259, 2266 (2014). The
Halberstam inquiry requires that “(1) the party the
defendant aids must perform a wrongful act that
causes an injury; (2) the defendant must be generally
aware of his role as part of an overall illegal or tortious
activity at the time that he provides the assistance;
and (3) the defendant must knowingly and
substantially assist the principal violation.”
Halberstam v. Welch, 705 F.2d 472, 487—88 (D.C. Cir.
1983).3 The Ninth Circuit interpreted the third of
these prongs to allow liability for a platform based
only on “general awareness” that a handful of users
might be using the platform in ways that support
offline terrorist conduct. If that is a sufficient basis for
liability, intermediaries will no longer be able to
2
This brief uses the term “intermediary” to refer to both
distributors and publishers of speech, as those two roles are
functionally the same in the context of online speech.
3
In the Justice Against Sponsors of Terrorism Act, which
amended the ATA in 2016, Congress instructed that Halberstam
provides the “‘proper legal framework for how [aiding-andabetting] liability should function’” under the ATA. App. 175a;
see 18 U.S.C. § 2333 Statutory Note (Findings and Purpose § 5)
(quoting Pub. L. No. 114—222, § 2, 130 Stat. 852 (2016)).
5
function as fora for others’ speech, and free expression
will be the loser.
This Court has long taken seriously the impact of
imposing overly broad indirect liability on speech
intermediaries. In a line of cases dating back to Smith
v. California, 361 U.S. 147 (1959), the Court has made
clear that laws that incentivize intermediaries’ selfcensorship through an insufficiently stringent
scienter requirement undermine the public’s ability to
engage in constitutionally protected speech and
accordingly violate the First Amendment. Smith and
its progeny concerned state laws that–by threatening
liability for facilitating others’ speech–prompted
intermediaries such as bookstores and newspapers to
curtail their distribution of protected speech. As this
Court emphasized, intermediaries’ rational caution in
the face of potential liability deprives the public of
robust access to all kinds of speech–exactly what the
First Amendment is designed to protect. Applying
similar principles here, the Court should require an
intermediary’s actual knowledge that a specific piece
of user-generated content on its platform provides
substantial assistance to a terrorist act before
imposing aiding-and-abetting liability.
If, instead, the Ninth Circuit’s startlingly broad
construction of the ATA stands, online intermediaries
will be forced to suppress protected speech, just like
the publishers and content distributors at issue in
Smith and its progeny. By allowing for indirect
liability based on mere “generalized knowledge” that
alleged terrorists or their affiliates use such
intermediaries’ services, this construction would
effectively require platforms to sharply limit the
content they allow users to post, lest courts find they
failed to take sufficiently “meaningful steps” against
6
speech later deemed beneficial to an organization
labeled “terrorist.”
Indeed, given the vast amounts of speech that
online intermediaries handle every day, were the
Ninth Circuit’s approach the law, intermediaries
would be likely to use necessarily blunt content
moderation tools to over-restrict speech or to impose
blanket bans on certain topics, speakers, or specific
types of content. Even today, online intermediaries
frequently take down content mistakenly identified as
offensive or forbidden–for instance, by confusing a
post about a landmark mosque with one about a
terrorist group.
Faced with potential ATA liability, all manner of
speech intermediaries–not only online platforms–
will grow more risk-averse and more susceptible to
overly cautious moderation, thus suppressing large
amounts of protected speech. And such a reading
would open the door to future federal or state
legislation imposing liability on online intermediaries
for inadvertently hosting other kinds of content,
thereby chilling platforms and inhibiting the public’s
access to speech on other topics. This problem is
exactly what the Court cautioned against in Smith
and its progeny.
This Court should accord proper respect to the
key First Amendment principles at stake and reverse
the Ninth Circuit’s erroneous interpretation of the
ATA.
7
ARGUMENT
I.
Interpretation of the Anti-Terrorism Act Must
Account for the Statute’s Effects on Speech
The First Amendment guards against the
possibility that the government will directly or
indirectly cause an intermediary of speech–whether
a bookstore or a social media platform–to suppress
otherwise protected speech of its users, including
speech about controversial topics like terrorism, for
fear of liability. This Court has long recognized that
stringent scienter requirements help avoid this
outcome and best serve First Amendment interests.
Consistent with this constitutional principle, the
scienter element of the multifactor Halberstam test–
the test that Congress directed should govern the
present inquiry–can and should be interpreted to
avoid the chilling effects that imposing overly
expansive indirect liability would have on the First
Amendment rights of intermediaries and their users.
In interpreting the ATA, this Court should avoid the
grave First Amendment harms that would result if
intermediaries were forced to engage in the sweeping
self-censorship and suppression of users’ lawful
speech that the Ninth Circuit’s extraordinarily
expansive reading of “knowing” assistance will
require.
A. The First Amendment Requires Robust
Protection for Intermediaries of Speech
Throughout its First Amendment jurisprudence,
this Court has consistently sought to avoid chilling
protected speech, including by imposing stringent
scienter requirements on statutes creating liability for
speech. See, e.g., Brandenburg v. Ohio, 395 U.S. 444,
448—49 (1969) (per curiam) (requiring intent to incite
8
or produce imminent lawless action for challenges to
inflammatory statements); New York Times Co. v.
Sullivan, 376 U.S. 254, 279—82 (1964) (requiring
“actual malice” for defamation claims by public
officials). That is particularly true where, as here,
laws potentially impose liability on intermediaries of
speech. Imposing liability on a speech intermediary
with an insufficiently stringent scienter requirement,
the Court has repeatedly warned, has a chilling effect
that limits both the intermediaries’ and the public’s
speech rights.
In Smith v. California, 361 U.S. 147 (1959), for
example, this Court invalidated a Los Angeles
ordinance imposing strict criminal liability on a
bookstore that sold obscene books.
The Court
explained that imposing strict liability would lead
booksellers to sell the limited number of books they
could actually inspect:
By dispensing with any requirement of
knowledge of the contents of the book on
the part of the seller, the ordinance tends
to impose a severe limitation on the
public’s
access
to
constitutionally
protected matter. For if the bookseller is
criminally liable without knowledge of the
contents, and the ordinance fulfills its
purpose, he will tend to restrict the books
he sells to those he has inspected; and thus
the State will have imposed a restriction
upon the distribution of constitutionally
protected as well as obscene literature.
Id. at 153. In that scenario, “[e]very bookseller would
be placed under an obligation to make himself aware
of the contents of every book in his shop,” and “[i]t
9
would be altogether unreasonable to demand so near
an approach to omniscience.” Id. (quotation marks
omitted). As a result, fewer books would be available
to the public: “If the contents of bookshops and
periodical stands were restricted to material of which
their proprietors had made an inspection, they might
be depleted indeed.” Id. This “self-censorship,”
compelled by the ordinance’s strict liability standard,
would “affect[] the whole public . . . . Through it, the
distribution of all books, both obscene and not obscene,
would be impeded.” Id. at 154. Notably, the Court
deemed the statute invalid because it lacked a
sufficient scienter requirement. See id.
The Court reaffirmed these principles in
Manual Enterprises, Inc. v. Day, 370 U.S. 478 (1962),
where a plurality held that magazine publishers could
not be liable under the Comstock Act for distributing
advertisements by independent photographers
offering nudist photographs for sale without proof that
the publisher “knew that at least some of his
advertisers were offering to sell obscene material.” Id.
at 492 (plurality op.) (emphasis added). The Court
explained:
Since publishers cannot practicably be
expected to investigate each of their
advertisers, and since the economic
consequences of an order barring even a
single issue of a periodical from the mails
might entail heavy financial sacrifice, a
magazine publisher might refrain from
accepting advertisements from those
whose own materials could conceivably be
deemed objectionable by the Post Office
Department. This would deprive such
10
materials, which might otherwise be
entitled to constitutional protection, of a
legitimate and recognized avenue of access
to the public.
Id. at 493.
Liability without proof of specific
knowledge of the character of the advertisements
included in the magazine “would as effectively ‘impose
a severe limitation on the public’s access to
constitutionally protected matter,’ . . . as would a state
obscenity statute which makes criminal the
possession of obscene material without proof of
scienter.” Id. at 492—93 (quoting Smith, 361 U.S. at
153).
Similarly, in Bantam Books, Inc. v. Sullivan, 372
U.S. 58 (1963), this Court held that a Rhode Island
commission violated the First Amendment by
threatening book distributors with liability for selling,
distributing, or displaying books to youth under the
age of 18 that the commission had deemed
“objectionable.” Id. at 61, 63—64. While the First
Amendment challenge in Bantam Books was brought
by book publishers, this Court explained that the
commission violated the Constitution by threatening
book distributors. See id. at 61. The Court held that
the Constitution requires “that regulation by the
States of obscenity conform to procedures that will
ensure against the curtailment of constitutionally
protected expression, which is often separated from
obscenity only by a dim and uncertain line.” Id. at 66.
And although the state commission in Bantam Books
had not seized or banned any books, the commission’s
“informal sanctions–the threat of invoking legal
sanctions and other means of coercion, persuasion,
and intimidation” resulted in the suppression of
11
constitutionally protected speech by making book
distributors unwilling to distribute the books. Id. at
67.4 As a result, both minors and adults were
“deprived of the opportunity to publish publications in
the State” that were not actually obscene. As the
Court explained:
[A]lthough the Commission’s supposed
concern is limited to youthful readers, the
‘cooperation’ it seeks from distributors
invariably
entails
the
complete
suppression of the listed publications;
adult readers are equally deprived of the
opportunity to purchase the publications
in the State.
Id. at 69 n.9.
In the years following these decisions, the Court
continued to recognize the importance of a heightened
scienter requirement before speech intermediaries
may be subjected to liability. See, e.g., Hamling v.
United States, 418 U.S. 87, 123 (1974) (holding that
obscenity prosecution against brochure distributors
was not unlawful because statute applied only to
“knowing” conduct); Ginsberg v. New York, 390 U.S.
629, 644 (1968) (upholding New York obscenity
statute and explaining that its scienter requirement
“rests on the necessity to avoid the hazard of selfcensorship of constitutionally protected material and
to compensate for the ambiguities inherent in the
4
The notices in Bantam Books caused book distributors “(a) to
refuse to take new orders for the proscribed publications, (b) to
cease selling any of the copies on hand, (c) to withdraw from
retailers all unsold copies, and (d) to return all unsold copies to
the publishers.” Bantam Books, 372 U.S. at 64.
12
definition of obscenity” (internal quotation omitted));
see also Seth F. Kreimer, Censorship by Proxy: The
First Amendment, Internet Intermediaries, and the
Problem of the Weakest Link, 155 U. Pa. L. Rev. 11,
83 (2006) (collecting cases and concluding that “[i]n
the last half century, Smith has regularly served as
the basis for decisions rejecting the imposition of
liability without fault on intermediaries who facilitate
the transmission of erotic materials from speaker to
listener”).5
Since the advent of the Internet, relatively few
cases have addressed these First Amendment
principles in the context of online speech, largely
5
The Court has likewise emphasized avoiding First
Amendment concerns when applying statutes of general
applicability that, like the ATA, might implicate speech. See,
e.g., Cohen v. California, 403 U.S. 15, 26 (1971) (invalidating
“breach of the peace” ordinance applied to forbid wearing a “Fuck
the Draft” jacket in a municipal courthouse, as an asserted
interest in avoiding public disturbances would “also run[] a
substantial risk of suppressing ideas in the process”); City of
Houston v. Hill, 482 U.S. 451, 453, 466 (1987) (holding that an
ordinance “that makes it unlawful to interrupt a police officer in
the performance of his or her duties” was unconstitutional, as it
“criminalizes a substantial amount of constitutionally protected
speech”); McCullen v. Coakley, 573 U.S. 464, 469, 476, 496—97
(2014) (invalidating on First Amendment grounds a
Massachusetts statute that made it a crime to knowingly stand
on a “public way or sidewalk” near certain abortion clinics, “even
though the Act says nothing about speech on its face,” because it
impermissibly limited petitioners’ ability “to converse with their
fellow citizens about an important subject on the public streets
and sidewalks”); see also Scales v. United States, 367 U.S. 203,
222 (1961) (construing statute to limit liability for membership
in Communist Party to only “‘active’ members” in part “because
of the close constitutional questions” that would arise from
imposing liability on “mere passive members”).
13
because Section 230 of the Communications Decency
Act, 47 U.S.C. § 230, has provided significant
protection from suits challenging platforms’
publication and dissemination of user-generated
content. But courts that have considered these
principles in the online context readily recognized that
the First Amendment precedents governing speech
intermediaries offer crucial guidance in determining
whether electronic services are liable for material
their users provided. In Cubby v. CompuServe, 776 F.
Supp. 135 (S.D.N.Y. 1991), for example, a plaintiff
sued the distributor of an electronic newsletter called
Rumorville for libel after the newsletter published
articles criticizing the plaintiff. See id. at 137—38. The
court first noted that “[t]he requirement that a
distributor must have knowledge of the contents of a
publication before liability can be imposed for
distributing that publication is deeply rooted in the
First Amendment,” citing Smith and its progeny. Id.
at 139—40. And while recognizing that the online
context was new, the court held that “it would be no
more feasible for CompuServe to examine every
publication it carries for potentially defamatory
statements than it would be for any other distributor
to do so.” Id. at 140. The court thus applied the Smith
rationale to CompuServe, explaining that “[a]
computerized database is the functional equivalent of
a more traditional news vendor, and the inconsistent
application of a lower standard of liability to an
electronic news distributor . . . than that which is
applied to a public library, book store, or newsstand
would impose an undue burden on the free flow of
information.” Id.
Similarly, in Backpage.com, LLC v. McKenna,
881 F. Supp. 2d 1262 (W.D. Wash. 2012), a district
14
court held that a Washington statute prohibiting the
“knowing” dissemination of advertisements for
commercial sex with a minor likely violated the First
Amendment under Smith. The court explained that
the statute, like the ordinance in Smith, “would
compel those publishers and distributors who did not
abstain from publishing large categories of speech
altogether to review every book, magazine, video, or
online post containing a ‘depiction’ and a possible
‘implicit’ ad for sex to ensure that none ran afoul of the
law”–i.e., to over-moderate and chill speech.
McKenna, 881 F. Supp. 2d at 1278 (quoting Smith, 361
U.S. at 153—54). Such a “pre-screening mechanism”
would “limit the amount of content available on some
publishers’ websites to the amount of content that
such publishers had the time and money to screen.”
Id. For example, “[s]ome individuals would be reticent
to provide government identification in connection
with borderline content, such as racy personal ads,
thus further diminishing the universe of protected
speech available online.” Id. As McKenna explained,
the “Constitution does not permit such collateral
burdens on protected speech.” Id.; cf. Universal
Comm’cn Sys., Inc. v. Lycos, 478 F.3d 413, 423 (1st Cir.
2007) (holding that “because of the serious First
Amendment issues that would be raised by allowing”
a state-law trademark claim against a message board
relating to comments posted by users, “the claim
would not survive, even in the absence of Section
230”).
In passing Section 230 in 1996, Congress sought
to provide a statutory shield for online intermediaries
involved in publication and dissemination of user
speech. But the core principles this Court articulated
decades earlier regarding the activities of traditional
15
publishers and distributors of speech–bookstores or
newspapers–apply with equal force to online
intermediaries today. As discussed below, holding the
platforms sued here liable under the ATA their users’
speech–without a robust scienter requirement–
would necessarily restrict the speech of its hundreds
of millions of users in violation of the First
Amendment principles enshrined in this Court’s
jurisprudence.
B. Where Aiding-and-Abetting Liability Under
the Anti-Terrorism Act Is Premised on
Speech, the Act Should Be Interpreted to
Avoid Chilling the Protected Speech of Online
Platforms and Their Users
Consistent with longstanding First Amendment
principles, the best reading of the ATA is to apply the
Halberstam framework to require more than
“generalized awareness” in cases like this one, where
liability is predicated on a defendant’s function as a
speech intermediary. In such cases, the best reading
of the ATA requires a showing that the intermediary
had actual knowledge that a specific piece of usergenerated content substantially assists an act of
terrorism. Congress directed courts to apply the ATA
using the Halberstam framework, which makes clear
that the scienter required in a particular case turns
critically on the “assistance” a defendant allegedly
provides. See Halberstam, 705 F.2d at 487—88.
Determining whether a defendant “knowingly
and substantially” assists in wrongful activity–
Halberstam’s third factor–requires consideration of
“the amount and kind of assistance.” Halberstam, 705
F.2d at 483 (cleaned up). Critical to this consideration
is the fact that Twitter, Facebook, and Google are
16
speech intermediaries.
Here, their alleged
“assistance” consists of operating mostly open fora for
large numbers of private speakers–including
disseminating and suggesting content to users based
on algorithms that seek to match users’ interests. But
just as a physical bookstore or newsstand does not
provide “substantial assistance” to an act of terrorism,
neither does operating a broadly accessible technology
platform without actual knowledge that third-party
speech is providing substantial assistance to an act of
terrorism. Thus, where, as here, indirect ATA liability
is predicated on speech carried by an intermediary,
the Court should–consistent with the ATA’s plain
text–read the Halberstam factor requiring “knowing
and substantial assistance” to mandate actual
knowledge that a particular post provides substantial
assistance to an act of terrorism. Otherwise, all sorts
of protected speech by all sorts of speakers could
conceivably subject an online intermediary to liability,
and intermediaries will have to restrict a great deal of
speech in order to mitigate the risk of litigation and
massive damages awards. This chilling effect, in turn,
will deprive the public of access to large swaths of
speech. These outcomes are precisely what the First
Amendment forbids under this Court’s precedents.
See supra Section I.A.
For example, under the Ninth Circuit’s
interpretation, a user’s post criticizing Israel’s actions
in the West Bank could subject an online intermediary
to liability, provided that the intermediary is
generally aware that the user is affiliated with a group
that has engaged in terrorist activities. By the same
token, a platform could be liable for failing to take
down user-generated content criticizing detention
policies at Guantanamo Bay, again on the ground that
17
those users (or the groups they belong to) are generally
known to engage in terrorist activities. The same
could be said of declining to moderate posts shared by
certain European militant organizations that support
specific candidates for office in the United States.
These types of speech fall squarely within the First
Amendment’s protections, yet an online intermediary
could face potentially crippling liability by failing to
delete them.
To avoid these chilling effects on the First
Amendment rights of intermediaries and their users
in accordance with this Court’s precedents, the Court
should read the ATA to require more than a showing
that an online platform had “general knowledge” that
some users associated with a terrorist organization
were among the hundreds of millions using its
platform. Instead, when ATA liability is predicated on
serving as an intermediary for another’s speech, the
Court should require actual knowledge that a specific
piece of user-generated content provided substantial
assistance to an act of terrorism.6
6
In other contexts, such as news reporting, courts have
recognized that an even more demanding scienter requirement
may be appropriate. See Rice v. Paladin Enters., Inc., 128 F.3d
233, 266 (4th Cir. 1997) (“News reporting, we can assume, no
matter how explicit it is in its description or depiction of criminal
activity, could never serve as a basis for aiding and abetting
liability consistent with the First Amendment.”).
18
II. The Ninth Circuit’s Overly Broad Interpretation
of ATA Liability as Applied to Speech Will Result
in the Suppression of Lawful Speech
If the Court accepts the Ninth Circuit’s
interpretation of the ATA, intermediaries–and
especially online intermediaries–will be compelled to
take extreme and speech-chilling steps to insulate
themselves from potential liability.
Given the
Internet’s enormity and the pervasiveness of content
associated with entities that might be labeled as
terrorists or associated with others who are, virtually
all online intermediaries could be said to possess the
“generalized knowledge” of such content that the
Ninth Circuit deemed sufficient to establish scienter
under the ATA. Under this interpretation, even
traditional publishers arguably have generalized
knowledge that coverage of the newsworthy activities
of terrorists, like coverage of any other newsworthy
activity, publicizes that activity. See, e.g., Interview
with Osama Bin Laden, PBS Frontline (May 1998),
https://perma.cc/9V9Y-UY9T;
Provisional
I.R.A.
Continues Truce, N.Y. Times, Jan. 3, 1975,
https://tinyurl.com/vkat5zsb (reporting on activities of
Provisional Irish Republican Army and quoting
extensively from a Provisional I.R.A. statement).
Indeed, and perversely, many Internet platforms may
obtain this knowledge by participating in initiatives
like the Global Internet Forum to Counter Terrorism,
an organization “committed to cross-industry efforts to
counter the spread of terrorist and violent extremist
content online.” About, Global Internet Forum to
Counter Terrorism, https://gifct.org/about/ (last
visited Dec. 3, 2022).
19
With “generalized knowledge” so easy to allege,
online intermediaries would be forced to conduct
overly aggressive content moderation to defend
against claims that they have “knowingly and
substantially assisted” a terrorist act by “refus[ing] to
take meaningful steps to prevent” their platforms’ use
by alleged terrorists. Gonzalez v. Google LLC, 2 F.4th
871, 908—09 (9th Cir. 2021). And because of the scale
of speech that many online platforms carry, content
moderation can be done only with necessarily blunt
instruments.
The result would be widespread
suppression of constitutionally protected speech on
some of the most important fora for public debate,
dialogue, and information today.
A. The Ninth Circuit’s Construction Incentivizes
Intermediaries to Over-Moderate Content,
Causing Dramatic Suppression of Protected
Speech
As in Smith, an online intermediary may “tend to
restrict the [content] he [hosts] to those he has
inspected; and thus the State will have imposed a
restriction upon the distribution of constitutionally
protected” content. Smith, 361 U.S. at 153. But the
Internet’s vast scale compounds the risk that the
Ninth Circuit’s overly expansive interpretation of ATA
liability will suppress lawful speech far beyond
anything this Court envisioned when it decided Smith.
Internet users create enormous amounts of content,
much of it posted through online intermediaries like
Twitter, Facebook, or Google. To give just one
20
snapshot: there are reportedly 500 million posts on
Twitter every day.7
Platforms threatened with liability based on
mere “generalized awareness” may well be forced to
restrict user-generated content to only the tiny
fraction that humans can review, excluding large
swaths of constitutionally protected content from their
services. Or intermediaries may prohibit certain
speakers altogether–on the ground that some might
argue their content provides assistance to terrorism–
and privilege others. For instance, an intermediary
might afford more latitude to favored or “trusted”
speakers like government officials or certain news
organizations. This too would chill the public’s access
to individual voices and diverse perspectives.
Or, as a third and perhaps the most likely
alternative, intermediaries may configure their
content moderation algorithms to restrict users’
speech that potentially falls within necessarily
overbroad definitions of objectionable content. Given
their scale, online intermediaries already rely
extensively on automated tools to promote compliance
with their content-moderation policies. See Hannah
Bloch-Wehba, Automation in Moderation, 53 Cornell
Int’l L.J. 41, 42, 48 (2020). But the Ninth Circuit’s
interpretation of the ATA will encourage
intermediaries to rely on automated content
moderation tools in a manner that will over-restrict
speech.
7
Jack Shepherd, 22 Essential Twitter Statistics You Need to
in 2022, Social Shepherd (Nov. 5, 2022),
https://tinyurl.com/5bjy4jw7.
Know
21
These
automated
content
moderation
technologies have inherent limits, which will
necessarily lead platforms to restrict more speech
than necessary in order to avoid the expansive liability
the Ninth Circuit rule would impose. To take one
example, platforms calibrate machine-learning tools
to filter content based on “confidence intervals,” a
measurement of the software’s confidence that it has
See
correctly identified forbidden content.
Automation in Moderation, supra, at 42, 48; Spandana
Singh, Everything in Moderation: An Analysis of How
Internet Platforms Are Using Artificial Intelligence to
Moderate User Generated Content, New America 5, 7
(July
15,
2019),
available
at
https://tinyurl.com/23zhxmaa; see also, e.g., Nafia
Chowdhury, Automated Content Moderation: A
Primer, Stanford Freeman Spogli Institute, 5 (March
19, 2022), https://tinyurl.com/3txpeufd (discussing
platforms’ use of confidence intervals). Moderation
tools set to a high confidence interval will block
content or flag it for human review only when there is
a high likelihood that it has correctly identified that
content, while setting such tools to a lower confidence
interval means that they will block or flag content
even when comparatively less sure that the content is
prohibited. A platform confronted with the Ninth
Circuit’s reading of the ATA might well set its
moderation system to reject users’ posts with only a
low confidence level that the material is associated
with or might be construed to assist alleged terrorist
organizations. And that will have the effect of
suppressing a vast amount of protected speech.
These concerns are not hypothetical. Machinelearning technology’s limitations, along with its
difficulty in discerning linguistic, cultural, or
22
historical context, already leads it to misidentify
benign content as harmful. See Carey Shenkman et
al., Do You See What I See? Capabilities and Limits of
Automated Multimedia Content Analysis, Center for
Democracy & Technology 27—29 (May 2021),
https://tinyurl.com/yzw2hmah.
One
such
“enforcement error,” for example, led Instagram to
remove a series of user-generated posts about the AlAqsa Mosque–one of Islam’s holiest sites–because
the term “al-Aqsa” also appears in the name of a
designated terrorist organization.
Jon Porter,
Instagram blames “enforcement error’ for removal of
posts about Al-Aqsa Mosque, The Verge (May 13,
2021), https://tinyurl.com/yn8kujej.
Similarly,
YouTube removed videos posted by independent
journalists showing the January 6, 2021 attack on the
U.S. Capitol after its content moderation system
determined the videos violated YouTube’s policies
against scams, deception, and spam. Mikael Thalen,
YouTube is cracking down on independent journalists
who covered the Capitol riot, Daily Dot (Feb. 3, 2021),
https://tinyurl.com/3kdwbuhx.
And
reportedly suspended dozens of Middle Eastern
journalists after potentially “miscategorizing their
accounts as having links to terrorism.” Olivia Solon,
“Facebook doesn’t care”: Activists say accounts
removed despite Zuckerberg's free-speech stance,
NBC
News
(June
https://tinyurl.com/4ve5nypm.
15,
2020),
Constitutionally protected speech like that
discussed above already is threatened by imperfect
automated content moderation, and the Ninth
Circuit’s reading of the ATA will vastly exacerbate the
problem.
23
B. The Ninth Circuit’s Construction Incentivizes
Intermediaries to Impose Blanket Bans on
Controversial Topics Like Terrorism, Barring
Speech Regardless of Its Constitutional
Protection
Online intermediaries may well be driven to
impose categorical limits on the kind of content they
host in response to the Ninth Circuit’s overly broad
interpretation of the ATA.
For example,
intermediaries
may
impose
across-the-board
restrictions on content with links to particular topics
or regions–including art, news reports, or even antiindoctrination materials–leaving speakers “deprived
of the opportunity” to publish constitutionally
protected content. Bantam Books, 372 U.S. at 71.
This outcome is especially likely for smaller platforms
without the resources to spend on sophisticated
automated moderation tools or armies of human
reviewers.
Again, experience teaches that these kinds of
categorical bans may be an attractive option to
intermediaries. Faced with potential legal exposure
under the Allow States and Victims to Fight Online
Sex Trafficking Act/Stop Enabling Sex Traffickers Act,
Pub. L. No. 115-164, 132 Stat. 1253 (2018) (“FOSTA”),
many platforms responded by removing or limiting the
availability of constitutionally protected content far
outside FOSTA’s ambit. For instance, Instagram
began removing content posted by authors writing
about sex work and even content related to sex
education. Abigail Moss, “Such a Backwards Step”:
Instagram Is Now Censoring Sex Education Accounts,
Vice (Jan. 8, 2021), https://tinyurl.com/5a4mezp9.
Tumblr took a similar step, announcing that “any
explicit posts will be flagged and deleted by
24
algorithms.” Shannon Liao, Tumblr will ban all adult
content on December 17th, The Verge (Dec. 3, 2018),
https://bit.ly/2SmoC5A.
With the Ninth Circuit’s rule, one could easily
imagine similar bans on all content even conceivably
related to terrorism. Such a prohibition might reach,
for example, a religious leader’s teachings against
violent extremism, a human rights organization’s
work documenting abuses perpetrated by terrorists, or
any number of other forms of protected speech. See,
e.g., Pope Francis (@pontifex), Twitter (Sept. 15,
2022), https://tinyurl.com/3ycm2hwd (“Extremism,
radicalism, terrorism and all other incentives to
hatred, hostility, violence and war have nothing to do
with the authentic spirit of religion and must be
rejected in the most decisive terms possible.”); Human
Rights Watch (@hrw), Twitter (Nov. 26, 2022),
https://tinyurl.com/ywa977cf (reporting that “[m]any
children repatriated from detention camps for ISIS
suspects and their families are successfully
reintegrating in their home countries”).
*
*
*
No matter how platforms respond under the
Ninth Circuit’s erroneous reading of the ATA–
limiting content by human review, requiring
“preclearance” of trusted speakers, relying on
imperfect and overinclusive automated moderation
tools, or imposing categorical bans on speech–free
speech will be the loser.
The Ninth Circuit’s
misinterpretation will lead intermediaries to take
steps that impede the distribution of all usergenerated
content,
including
constitutionally
protected and socially beneficial speech by persons
with no tie to anyone who could be deemed a terrorist.
25
A proper interpretation of the ATA–where, as here,
liability is predicated on acts of speech–avoids these
significant First Amendment problems. This Court
should reject the Ninth Circuit’s view and its
implications for protected online speech of all kinds.
CONCLUSION
For the foregoing reasons, this Court should
reverse the Ninth Circuit’s decision below.
Respectfully submitted,
Samir Jain
Emma Llansó
Caitlin Vogus
CENTER FOR DEMOCRACY
& TECHNOLOGY
1401 K Street NW, Suite
200
Washington, DC 20005
David D. Cole
AMERICAN CIVIL
LIBERTIES UNION
FOUNDATION
915 15th Street NW
Washington, DC 20005
Kathleen R. Hartnett
Counsel of Record
Benjamin Kleine
Reece Trevor
COOLEY LLP
3 Embarcadero Center
20th Floor
San Francisco, CA 94111
(415) 693-2000
khartnett@cooley.com
Patrick J. Hayden
COOLEY LLP
55 Hudson Yards
New York, NY 10001
Jennifer Stisa Granick
AMERICAN CIVIL
LIBERTIES UNION
FOUNDATION
39 Drumm Street
San Francisco, CA 94111
Aaron Mackey
ELECTRONIC FRONTIER
FOUNDATION
815 Eddy Street
San Francisco, CA 94109
Patrick Toomey
Sarah Taitz
Bruce D. Brown
Katie Townsend
Elizabeth Gyori
AMERICAN CIVIL
LIBERTIES UNION
FOUNDATION
125 Broad Street
New York, NY 10004
Nicole A. Ozer
AMERICAN CIVIL
LIBERTIES UNION
FOUNDATION OF
NORTHERN CALIFORNIA
39 Drumm Street
San Francisco, CA 94111
Chris Riley
Jonathan Cannon
THE R STREET INSTITUTE
1212 New York Avenue
NW,
Suite 900
Washington, DC 20005
Counsel for Amici Curiae
December 5, 2022
26
Gabe Rottman
Grayson Clary
REPORTERS COMMITTEE
FOR FREEDOM OF THE
PRESS
1156 15th Street NW,
Suite 1020
Washington, DC 20005
Alex Abdo
Anna Diakun
Alexia Ramirez
KNIGHT FIRST
AMENDMENT INSTITUTE
AT COLUMBIA
UNIVERSITY
475 Riverside Drive,
Suite 302
New York, NY 10115
This is a copy of a public record, reproduced as it was published. It is not legal advice, and it may not be the version a court would rely on. Check the official source before you cite it.