Doe I v. Cisco Systems, Inc.

               FOR PUBLICATION

  UNITED STATES COURT OF APPEALS
       FOR THE NINTH CIRCUIT

DOE I; DOE II; IVY HE; DOE III;          No. 15-16909
DOE IV; DOE V; DOE VI;
CHARLES LEE; ROE VII; ROE VIII;      D.C. No. 5:11-cvLIU GUIFU; DOE IX; WEIYU                02449-EJD
WANG, and those individuals
similarly situated,
                                           ORDER
            Plaintiffs-Appellants,

 v.

CISCO SYSTEMS, INC.; JOHN
CHAMBERS; FREDY CHEUNG,
AKA Zhang Sihua; DOES, 1-100,

            Defendants-Appellees.

               Filed September 3, 2024

  Before: A. Wallace Tashima, Marsha S. Berzon, and
           Morgan Christen, Circuit Judges.

                       Order;
             Statement by Judge Berzon;
             Dissent by Judge Bumatay
2                   DOE I V. CISCO SYSTEMS, INC.


                          SUMMARY *


                       Alien Tort Statute

    The panel denied a petition for panel rehearing and a
petition for rehearing en banc in a case in which the panel
affirmed in part and reversed in part the district court’s
judgment dismissing claims under the Alien Tort Statute
(“ATS”) and the Torture Victim Protection Act, and
remanded.
   The action was brought by practitioners of Falun Gong
who alleged that they or family members were victims of
human rights abuses by the Chinese Communist Party and
Chinese government officials and that these abuses were
enabled by technological assistance of U.S. corporation
Cisco Systems, Inc., and two Cisco executives.
    Respecting the denial of rehearing en banc, Judge
Berzon, joined by Judges Tashima and Paez, wrote that the
decision to deny rehearing en banc was correct because this
court is not free to depart from Sosa v. Alvarez-Machain, 

 (2004), which provides a two-step test for
determining whether a cause of action may lie under the
ATS. She wrote that the panel majority faithfully applied
the Sosa framework to the facts of this case.
    Judge Bumatay, joined by Judges Callahan, Ikuta,
Bennett, R. Nelson, and VanDyke, dissented from the denial
of rehearing en banc. He wrote that the court made three
main errors in refusing to reconsider the case en banc. First,

*
 This summary constitutes no part of the opinion of the court. It has
been prepared by court staff for the convenience of the reader.
                  DOE I V. CISCO SYSTEMS, INC.               3


the court failed to restrict ATS liability to causes of action
comparable to historically recognized torts. Second, the
court violated the separation of powers in pronouncing a new
cause of action—even though Congress has continued to
legislate in this very area. Third, the court ignored serious
foreign-policy concerns—permitting federal courts to
intrude in the delicate relations with another world
superpower.


                          ORDER

    Judge Tashima and Judge Berzon voted to deny the
petition for panel rehearing and recommended denying the
petition for rehearing en banc. Judge Christen voted to grant
the petition for panel rehearing and the petition for rehearing
en banc. A judge requested a vote on whether to rehear the
matter en banc. The matter failed to receive a majority of
the votes of the nonrecused active judges in favor of en banc
consideration. Fed. R. App. P. 35(a). Judges Wardlaw,
Nguyen, and Collins did not participate in the deliberations
or vote in this case.
   Appellees’ Petition for Panel Rehearing and Rehearing
En Banc, filed August 11, 2023, is DENIED.


BERZON, Circuit Judge, with whom TASHIMA and PAEZ,
Circuit Judges, join, respecting the denial of rehearing en
banc:

    The Alien Tort Statute (“ATS”) provides in full: “The
district courts shall have original jurisdiction of any civil
action by an alien for a tort only, committed in violation of
4                 DOE I V. CISCO SYSTEMS, INC.


the law of nations or a treaty of the United States.” 

. The ATS is “only jurisdictional” and does not itself
provide a cause of action. Sosa v. Alvarez-Machain, 

 (2004). How then should a court determine
what causes of action are cognizable as “violation[s] of the
law of nations?”
    Judge Bumatay offers one answer—those actions that
the First Congress had “in mind” when it enacted the ATS:
“violation of safe conducts, infringement of the rights of
ambassadors, and piracy.” Dissent from Denial of En Banc
at 15 (quoting Sosa, 

). Absent congressional
action, Judge Bumatay argues, judicial recognition of
additional causes of actions or forms of liability is
“extraordinarily disfavored if not dead letter.” Dissent from
Denial of En Banc at 25. That view has been championed
by several Supreme Court justices, but, critically, it has
never gained the support of a majority of the Court. Instead,
Sosa provides a different answer to this question than does
Judge Bumatay. As a court of appeals, we are not free to
depart from that binding precedent.
    Sosa provides a two-step test for determining whether a
cause of action may lie under the ATS. At step one, courts
ask whether the claim “rest[s] on a norm of international
character accepted by the civilized world and defined with a
specificity comparable to the features of” the torts
recognized to constitute a violation of the law of nations at
the time the ATS was enacted. Sosa, 

.
    At step two, ‘‘it must be determined further whether
allowing [a] case to proceed under the ATS is a proper
exercise of judicial discretion, or instead whether caution
requires the political branches to grant specific authority
before [a new form of liability] can be imposed.” Jesner v.
                  DOE I V. CISCO SYSTEMS, INC.               5


Arab Bank, PLC, 

 (2018). Step two
requires consideration of the foreign policy implications and
“practical consequences of making [a] cause available to
litigants in the federal courts.” Sosa, 542 U.S. at 732–33.
Additionally, courts must consider whether “there are sound
reasons to think Congress might doubt the efficacy or
necessity of a . . . remedy” before recognizing a new cause
of action. Jesner, 

 (quoting Ziglar v. Abbasi,

 (2017)).
    The panel majority faithfully applied the Sosa
framework to the facts of this case. Judge Bumatay does not
really take issue with the majority’s application of Sosa.
Instead, he recasts Sosa’s standard in a new mold. In service
of this endeavor, he relies largely on plurality opinions and
concurrences and invokes issues not raised by the parties,
violating the principle of party presentation, United States v.
Sineneng-Smith, 

, 375–76 (2020). Because this
Court is not free to depart from Sosa, the decision to deny
rehearing en banc was correct.
                               I
    The panel majority in this case began by considering
whether aiding and abetting liability is “sufficiently definite
and universal to be a viable form of liability under the ATS.”
Doe I v. Cisco Sys., Inc., 

 (9th Cir. 2023).
In answering this question, the panel majority looked to the
material Sosa instructs courts to consider: ‘‘those sources we
have long, albeit cautiously, recognized,’’ which include
treaties as well as “the customs and usages of civilized
nations; and, as evidence of these, . . . the works of
[qualified] jurists and commentators.” Sosa, 542 U.S. at
733–34 (quoting The Paquete Habana, 

(1900)). In doing so, the panel majority relied heavily on the
6                     DOE I V. CISCO SYSTEMS, INC.


careful reasoning of Judge Katzmann in Khulumani v.
Barclay National Bank Ltd., 

 (2d Cir. 2007),
who reviewed the decisions of several international
tribunals, multiple treaties and conventions, the actions of
the U.N. Security Council, and the Rome Statute of the
International Criminal Court.       504 F.3d at 270–77
(Katzmann, J., concurring). In Khulumani, Judge Katzmann
concluded that aiding and abetting liability was sufficiently
well defined and universally recognized under international
law to be cognizable under the ATS. 

.
     Khulumani and the panel majority’s application of Sosa
and recognition of aiding and abetting liability under the
ATS is in keeping with every other circuit to have addressed
the issue. See Aziz v. Alcolac, Inc., 

 (4th
Cir. 2011); Presbyterian Church of Sudan v. Talisman
Energy, Inc., 

, 258–59 (2d Cir. 2009); Romero
v. Drummond Co., 

 (11th Cir. 2008)
(citing Cabello v. Fernandez-Larios, 

, 1157–
58 (11th Cir. 2005)); cf. Doe v. Exxon Mobil Corp., 

 (D.C. Cir. 2011), vacated on other grounds, 

 (D.C. Cir. 2013). 1
   Further, nothing in Sosa or the circuit court opinions that
have held aiding and abetting liability cognizable under the
ATS suggests that the tort-by-tort analysis of aiding and
abetting liability that Judge Bumatay champions is required.
Importantly, Cisco has never argued for such an approach.


1
 Although she dissented from the panel majority’s analysis of aiding and
abetting liability under Sosa’s step two, Judge Christen stated that “[t]he
majority’s careful and cogent analysis of aiding and abetting liability
under the Alien Tort Statute . . . is consistent with the views of our sister
circuits, and in an appropriate case, I would likely join it.” Cisco, 

 (Christen, J., concurring in part and dissenting in part).
                   DOE I V. CISCO SYSTEMS, INC.                  7


In its motion to dismiss before the district court and in both
its answering brief and its petition for rehearing en banc
before this Court, Cisco has consistently argued that aiding
and abetting liability is generically unavailable under the
ATS, not that it is unavailable for the particular underlying
torts alleged in the operative complaint. Nor has Cisco ever
contested the availability of a cause of action under the ATS
for four of the plaintiffs’ seven alleged violations—torture,
prolonged arbitrary detention, disappearance, and
extrajudicial killing—if committed directly. The principle
of party presentation commands that we “rely on the parties
to frame the issues for decision and assign to courts the role
of neutral arbiter of matters the parties present.” Sineneng-Smith, 

 (quoting Greenlaw v. United States,

 (2008)).
    Notably, the panel majority’s generic approach is
consistent with that employed by the Supreme Court in
Jesner. The Jesner Court “assumed” that inflicting death or
injury by terrorism as well as “knowingly and purposefully
facilitat[ing] banking transactions to aid, enable, or
facilitate . . . terrorist acts” constituted “crimes in violation
of well-settled, fundamental precepts of international law”
before considering whether foreign corporations may be
liable under the ATS as a general matter. 

.
The panel majority took the same approach here, where
some of the underlying causes of action were unchallenged,
leaving it to the district court to “consider in the first instance
the viability of the substantive claims under the ATS to the
degree that viability is contested.” Cisco, 

.
                                II
   Judge Bumatay also offers a conception of the
separation-of-powers concerns included in Sosa’s second
8                DOE I V. CISCO SYSTEMS, INC.


step that contradicts governing law. Sosa requires that a
court consider whether Congress “might doubt the efficacy
or necessity of a . . . remedy” before recognizing a new cause
of action under the ATS. Jesner, 

 (quoting
Ziglar v. Abbasi, 

 (2017)). The panel
majority engaged in this exact inquiry. Cisco, 73 F.4th at
722–24. Contrary to Judge Bumatay’s argument, neither the
principles of Erie Railroad Co. v. Tompkins, 

(1938), nor the Supreme Court’s ATS jurisprudence postSosa command that the answer to this question “should
always be Congress.” See Dissent from Denial of En Banc
at 32 (emphasis in original).
    The Sosa majority explicitly considered and rejected
Justice Scalia’s contention in his partial concurrence that
Erie and the modern, “positivistic” conception of the
common law on which Erie relied had “close[d] the door to
further independent judicial recognition of actionable
international norms.” 

; see 

 at 744–46
(Scalia, J., concurring in part and concurring in judgment).
The Sosa majority explained:

       We think an attempt to justify such a position
       would be particularly unconvincing in light
       of what we know about congressional
       understanding bearing on this issue lying at
       the intersection of the judicial and legislative
       powers. The First Congress, which reflected
       the understanding of the framing generation
       and included some of the Framers, assumed
       that federal courts could properly identify
       some international norms as enforceable in
       the exercise of § 1350 jurisdiction. We think
       it would be unreasonable to assume that the
                 DOE I V. CISCO SYSTEMS, INC.               9


       First Congress would have expected federal
       courts to lose all capacity to recognize
       enforceable international norms simply
       because the common law might lose some
       metaphysical cachet on the road to modern
       realism. Later Congresses seem to have
       shared our view. The position we take today
       has been assumed by some federal courts for
       24 years, ever since the Second Circuit
       decided Filartiga v. Pena-Irala, 

       (C.A.2 1980), and for practical purposes the
       point of today’s disagreement has been
       focused since the exchange between Judge
       Edwards and Judge Bork in Tel-Oren v.
       Libyan Arab Republic, 

       (C.A.D.C. 1984). Congress, however, has
       not only expressed no disagreement with our
       view of the proper exercise of the judicial
       power, but has responded to its most notable
       instance     by       enacting     legislation
       supplementing the judicial determination in
       some detail.

 at 730–31 (citing id. at 728 (discussing the Torture
Victim Protection Act)).
    Sosa further considered and rejected Justice Scalia’s
contention in his concurrence that the ATS should be read to
“preclude federal courts from recognizing any further
international norms as judicially enforceable.” Id. at 729.
Instead, the Court held that claims under the ATS “must be
gauged against the current state of international law.” Id. at
733. In the twenty years since Sosa was decided, no majority
of the Supreme Court has ever held that “separation-of-
10                DOE I V. CISCO SYSTEMS, INC.


powers concerns almost entirely foreclose the recognition of
new causes of action under the ATS.” Dissent from Denial
of En Banc at 25.
    Kiobel v. Royal Dutch Petroleum Co., 

(2013), for example, held simply that the ATS does not apply
extraterritorially. 

. Nestlé USA, Inc. v. Doe,

 (2021), relied on Kiobel in holding that “[t]o
plead facts sufficient to support a domestic application of the
ATS, plaintiffs must allege more domestic conduct than
general corporate activity.” 

. Jesner held
that foreign corporations cannot be liable under the ATS.

. The majority opinion in Jesner considered
but again declined to adopt the argument that “a proper
application of Sosa would preclude courts from ever
recognizing any new causes of action under the ATS.” 

. This Court is not free to adopt a view twice rejected
by the Supreme Court in favor of one advocated for in
minority concurrences.
    Further, as the panel majority explained, Central Bank of
Denver, N.A. v. First Interstate Bank of Denver, N.A., 

 (1994), is inapposite here. See Cisco, 

. As Sosa makes clear, the scope of liability under the
ATS is determined by reference to international law not
federal statutes. See 

. Perhaps more to the
point for present purposes, Central Bank did not announce a
presumption against aiding and abetting liability; it simply
rejected a presumption favoring such liability. The panel
majority’s aiding and abetting analysis is not based on any
presumption of aiding and abetting liability. It is based on a
de novo consideration of well-developed international law
materials, applying no presumption one way or the other.
Central Bank’s rejection of a presumption of aiding and
abetting liability for federal civil law is for both these
                  DOE I V. CISCO SYSTEMS, INC.               11


reasons irrelevant to the question of whether such liability is
“sufficiently definite” and universal under international law.
See 

.
                              III
    Judge Bumatay’s invocation of foreign policy
considerations, relevant to Sosa’s step two, is similarly offbase. As the majority opinion explains, recognizing aiding
and abetting liability here “does not trigger Sosa’s principal
foreign policy concern—that ATS claims could impose
liability on sovereign nations for behavior with respect to
their own citizens.” Cisco, 

. Plaintiffs here
do not “seek to hold China and its government accountable
for purported violations of international law.” Dissent from
Denial of En Banc at 36. No claims were brought against
China or Chinese government officials. Claims for aiding
and abetting liability against a U.S. corporation do not
present the same foreign policy concerns as in Doe v. Qi, 

 (N.D. Cal. 2004), for example, where the
defendants were a former mayor of Beijing and a regional
governor. 349 F. Supp. 2d at 1267–68.
    The silence of the U.S. Department of State further
distinguishes the foreign policy implications here from those
in Doe v. Qi, where both the Department of State and the
Chinese government submitted statements of interest. See

 at 1296–97, 1300. The probative value of the United
States’ failure to comment on this case does not rest on an
expectation that the Department of State monitor the
activities of “all 94 federal district courts.” Dissent from
Denial of En Banc at 38 (quoting Cisco, 

(Christen, J., concurring in part and dissenting in part)). This
Court first heard oral argument in this case on April 18,
2017. Submission was then vacated while the Supreme
12                DOE I V. CISCO SYSTEMS, INC.


Court decided Jesner and Nestlé. We held oral argument
again after Nestlé was decided by the Supreme Court. At no
point in the seven years since that first oral argument has the
United States taken steps to advise this court of its views or
otherwise taken action regarding this case. And the United
States has remained silent even since Cisco first argued, in
its petition for rehearing en banc, that this Court should
solicit the views of the national government. Cisco was of
course free to ask the United States to chime in with its views
on the foreign policy implications of the case at the petition
for rehearing en banc stage; either it did not do so or the
government chose not to come forward.
     Thus, it is even clearer now than it was at the time the
opinion was filed that the foreign policy implications here
are not of sufficient concern to the United States government
to trigger its involvement at this juncture. Why that may be
we do not know. In any event, the panel made clear that on
remand, the District Court may request views from the
Department of State. Cisco, 

.
    Moreover, whether to ask the government for its views
is within the discretion of the panel. An en banc panel cannot
productively spend its energies attempting to cabin that
discretion, which is a subject more worthy (if it has any
worth at all) of a rules committee. It would have been foolish
in the extreme to have taken a case en banc for a panel’s
discretionary decision not to ask the government for its
views.
    Although Judge Bumatay takes a different view of the
foreign policy implications of this case, a difference of
opinion regarding prudential concerns as applied to
particular facts is an insufficient reason for en banc review.
                 DOE I V. CISCO SYSTEMS, INC.             13


                            …
    Sosa instructs that “[f]or two centuries we have affirmed
that the domestic law of the United States recognizes the law
of nations. It would take some explaining to say now that
federal courts must avert their gaze entirely from any
international norm intended to protect individuals.” 542
U.S. at 729–30 (citations omitted). Averting our gaze is just
what the dissent from denial of en banc would have us do.
But so long as Sosa remains good law, that choice is not ours
to make, and that explanation is not ours to provide. The
decision not to rehear Cisco en banc was correct.


BUMATAY, Circuit Judge, joined by CALLAHAN,
IKUTA, BENNETT, R. NELSON, and VANDYKE, Circuit
Judges, dissenting from the denial of rehearing en banc:

    In 1789, the First Congress passed the Alien Tort Statute
(“ATS”) to grant federal courts jurisdiction over tort suits
involving violations of the “law of nations.” Judiciary Act
of 1789, ch. 20, § 9, 

, 77. As a jurisdictional
provision, the ATS opened the federal courthouse doors to
victims of certain torts. But the statute didn’t create those
torts itself. Rather, the private right of action must come
from somewhere else. So in creating tort liability for
violations of international law, we must ask—who gets to
decide?
    Twenty years ago, Justice Scalia provided a simple
answer to this question: “We Americans have a method for
making the laws that are over us. We elect representatives
to two Houses of Congress, each of which must enact the
new law and present it for the approval of a President, whom
14                DOE I V. CISCO SYSTEMS, INC.


we also elect.” Sosa v. Alvarez-Machain, 

(2004) (Scalia, J., concurring). So in deciding what is “so
universally disapproved by other nations that it is
automatically unlawful” in this country and “automatically
gives rise to a private action for money damages in federal
courts,” we look to one source—“American law—the law
made by the people’s democratically                 elected
representatives.” 

.
    Since then, other Justices have expressed similar views.
See, e.g., Jesner v. Arab Bank, PLC, 

(2018) (Gorsuch, J., concurring) (“In our democracy the
people’s elected representatives make the laws that govern
them. Judges do not. The Constitution’s provisions
insulating judges from political accountability may promote
our ability to render impartial judgments in disputes between
the people, but they do nothing to recommend us as
policymakers for a large nation.”). So great is the need to
defer to the political branches that several Justices have
made clear, outside the “historical torts likely on the mind of
the First Congress,” “courts may not create a cause of action
for [other] torts.” Nestlé USA, Inc. v. Doe, 

(2021) (plurality) (Thomas, J., joined by Gorsuch and
Kavanaugh, JJ.) (emphasis added). Indeed, for more than
230 years, the Supreme Court has never “invoked the ATS
to create a new cause of action.” 

 (Gorsuch, J.,
concurring).
    But the Ninth Circuit gives a different answer to this
question. When asking who may create a new cause of
action for violations of international law, our answer is
merely—two judges of our court. Today, we do not revisit
the view of a divided panel of our court that recognizes a
brand-new type of liability—aiding and abetting—for any
tort claimed under a broad conception of international law.
                  DOE I V. CISCO SYSTEMS, INC.               15


Because the panel majority viewed this accomplice liability
to be “a norm of customary international law,” it felt free to
inaugurate a new cause of action under the ATS. See Doe I
v. Cisco Sys., Inc., 

 (9th Cir. 2023). The
panel majority justified its new creation by claiming that its
decision “does not raise separation-of-powers or foreign
policy concerns.” 

 With this decision, our circuit now
holds that aiding-and-abetting liability automatically
attaches to any tort claiming a violation of international law.
    Our answer is, of course, wrong. The ATS, as recodified, provides that “[t]he district courts shall have
original jurisdiction of any civil action by an alien for a tort
only, committed in violation of the law of nations or a treaty
of the United States.” 

. By its terms, it is
only about jurisdiction. The ATS does not confer broad
lawmaking powers on courts, and it says nothing about
making new judge-created causes of action. It merely allows
aliens to bring suits for certain torts recognized under the law
of nations in federal courts.
    Contrary to the panel majority’s view, the law of nations
isn’t a freewheeling concept. When it was enacted, it had a
definite and limited meaning—it referred to a “modest
number of international law violations” for which the
general “common law would provide a cause of action.”
Sosa, 

; see also Nestlé, 

(plurality) (observing that federal courts would recognize
torts “under general common law”). And it’s widely
assumed that the First Congress likely had only “three
primary offenses” “in mind” in enacting the ATS: “violation
of safe conducts, infringement of the rights of ambassadors,
and piracy.” Sosa, 

. True, both the concepts
of general common law and international law have changed
in the two-and-a-half centuries since its enactment. But that
16                DOE I V. CISCO SYSTEMS, INC.


doesn’t mean we have license to create any cause of action
so long as we deem it an “international norm.” Indeed, such
an expansive view might incorporate countless modern
international-law customs and norms into domestic law—all
enforceable in federal court. Instead, the Court has made
clear that we must not fashion a new cause of action if any
“sound reason[]” exists “to think Congress might doubt the
efficacy or necessity of [the new] remedy.” Jesner, 

 (simplified). In this case, sound reasons abound.
    We make three main errors in refusing to reconsider this
case en banc. First, we failed to restrict ATS liability to
causes of action comparable to historically recognized torts.
Second, we violated the separation of powers in pronouncing
a new cause of action—even though Congress has continued
to legislate in this very area. And third, we ignored serious
foreign-policy concerns—permitting federal courts to
intrude in the delicate relations with another world
superpower. All three reasons show that the Ninth Circuit
got it wrong here.
    First, we flouted the ATS’s historical understanding.
Relying on modern United Nations documents and other
international examples, the panel majority created universal
aiding-and-abetting liability for all torts claiming a violation
of international law. The panel majority proclaimed
accomplice liability an “international norm”—without
limitation—no matter the tort. But the panel majority didn’t
justify this sweeping ruling by examining the international
norms for each tort alleged. Nor did it attempt to reconcile
its views with the history of the ATS. Compared to the
historical understanding of the law, there’s reason to
question whether universal accomplice liability should be
part of our federal common law. See Cent. Bank of Denver,
N.A. v. First Interstate Bank of Denver, N.A., 

,
                 DOE I V. CISCO SYSTEMS, INC.             17


181 (1994) (holding that, while aiding and abetting is “an
ancient criminal law doctrine,” it is “at best uncertain in
application” for tort liability). So the panel majority was
wrong to claim that universal aiding-and-abetting liability
should apply under the ATS.
    Second, we violate the separation of powers. While
Congress knows how to delineate accomplice liability, it
didn’t provide for aiding and abetting in the ATS and we
can’t presume that Congress meant to include it. 

(“[W]hen Congress enacts a statute under which a person
may sue and recover damages from a private defendant for
the defendant’s violation of some statutory norm, there is no
general presumption that the plaintiff may also sue aiders
and abettors.”). And over the years, including in the ATS’s
re-codification, Congress has declined to amend the statute
to add accomplice liability. In fact, Congress has already
expanded the reach of the ATS—to recognize “torture” as a
tort through the Torture Victim Protection Act of 1991
(“TVPA”), 

 note. Absent a signal from
Congress to create aiding-and-abetting liability, doing so
now usurps the lawmaking role—a function the Constitution
gives only to the political branches.
    Third, we interfere in international affairs. When
conceived, the ATS’s purpose was to shield the country from
being drawn into disputes with other nations. See Jesner,

 (“The ATS was intended to promote
harmony in international relations by ensuring foreign
plaintiffs a remedy for international-law violations in
circumstances where the absence of such a remedy might
provoke foreign nations to hold the United States
accountable.”). Our court has now transformed this shield
into a sword to punish foreign nations through their alleged
18                    DOE I V. CISCO SYSTEMS, INC.


aiders and abettors. So rather than promote harmony, we
arm the ATS to incite international conflict.
    And this is a worrisome undertaking—with real world
consequences. Twenty years ago, the State Department
warned us about the foreign-policy implications of
confronting China in a case very much like this one. See
Letter from William H. Taft, IV, Dep’t of State, to Assistant
Att’y Gen. Robert D. McCallum, Re: Doe v. Qi (N.D. Cal.)
(Sep. 25, 2002) (“2002 State Dep’t Letter”). 1 It warned of
the concern for reciprocal actions against our citizens and
officials. 

 And there’s little reason to think this concern
has eased over the last 20 years. Indeed, the panel majority
refused to request the view of the United States, over Judge
Christen’s dissent. The likely reason why? The United
States has long opposed the creation of a cause of action for
aiding-and-abetting liability under the ATS. See Nestlé, 

. So our actions intrude into the executive
branch’s function, touching on foreign-policy questions that
we have no competence to understand.
    Because we overstepped our authority to create new
liability under the ATS, I respectfully dissent from the denial
of rehearing en banc.
                                   I.
                                   A.
    The ATS was passed by the First Congress as part of the
Judiciary Act of 1789. In this Act, Congress created the
federal courts and defined their jurisdiction. The original
text provided that federal courts:



1
    Available at: 2009-2017.state.gov/documents/organization/57535.pdf.
                  DOE I V. CISCO SYSTEMS, INC.               19


       shall also have cognizance, concurrent with
       the courts of the several States, or the circuit
       courts, as the case may be, of all causes where
       an alien sues for a tort only in violation of the
       law of nations or a treaty of the United States.

Judiciary Act of 1789, ch. 20, § 9, 

, 77. At the time,
the ATS was enacted to protect our fledgling nation from
international conflict arising from uncompensated violations
of the law of nations. By giving jurisdiction to federal courts
to redress violations of international law, the ATS ensured
our country’s compliance with the law of nations and
prevented other nations from seeking vindication against our
country through other means, such as war. See Anthony J.
Bellia, Jr. & Bradford R. Clark, The Alien Tort Statute and
the Law of Nations, 

, 448–49 (2011).
    The phrase “law of nations” had a distinct, narrow
meaning in 1789. The idea of the law of nations took root in
early European thought as political philosophers and legal
scholars sought to explain interactions between countries.
By 1789, the term “law of nations” had been extensively
discussed and explained by well-known legal scholars. See
4 William Blackstone, Commentaries *66–67 (1769) (“The
law of nations is a system of rules, deducible by natural
reason, and established by universal consent among the
civilized inhabitants of the world[.]”); Emmerich de Vattel,
Les Droit de gens [The Law of Nations], 30 (1797) (Kapossy
& Whatmore eds., 2011) (defining the “law of nations” as
“the science which teaches the rights subsisting between
nations or states, and the obligations correspondent to those
rights”). At its heart, the “law of nations” identified customs
that governed interactions between countries. See Anthony
J. Bellia, Jr., and Bradford R. Clark, The Federal Common
20                  DOE I V. CISCO SYSTEMS, INC.


Law of Nations, 

, 6 (2009) (“[A]
foundational principle of the law of nations” at the Founding
“was that each nation should reciprocally respect certain
perfect rights of every other nation to exercise territorial
sovereignty, conduct diplomatic relations, exercise neutral
rights, and peaceably enjoy liberty.”).
    When the ATS was passed, what torts violated the law
of nations was a matter of general law or “general common
law.” See Sosa, 

 (Scalia, J., concurring).
Under this understanding, “customary international law was
not regarded as federal law, but as a species of
nonpreemptive ‘general law.’” Bellia & Clark, The Federal
Common Law of Nations, 109 Colum. L. Rev. at 4. Even so,
scholarly debate persists about what “torts” would have been
considered “in violation of the law of nations” at the time of
the First Congress. Looking to Blackstone, the Supreme
Court assumed that the First Congress had only three
primary violations “in mind” when it enacted the ATS: the
violation of safe conducts, infringement of the rights of
ambassadors, and piracy. Sosa, 

. Some view
this grouping as too narrow. See Bellia & Clark, The Alien
Tort Statute and the Law of Nations, 78 U. Chi. L. Rev. at
454 (arguing that the ATS also permitted jurisdiction more
expansively for “common law tort claims by aliens against
United States citizens for intentional injuries to person or
property”). Others view it as too broad. See Thomas H. Lee,
The Safe-Conduct Theory of the Alien Tort Statute, 

, 836 (2006) (positing “that the ATS was
enacted [only] to allow aliens to sue . . . for transgressions of
safe conducts”). 2

2
  This model of granting federal courts jurisdiction over a substantive
area of law is not unique. For example, “[t]he Judiciary Act of 1789
                      DOE I V. CISCO SYSTEMS, INC.                          21


    In any case, following Erie Railroad Co. v. Tompkins,

 (1938), which marked the “death of the old
general common law,” development of torts under the law of
nations changed. Sosa, 

 (Scalia, J.,
concurring). Today, rather than discern general law, federal
courts may only develop “federal common law” that is, in
the words of Justice Scalia, “self-consciously ‘made’ rather
than ‘discovered.’” 

 Because of the nature of this judge-made law, “federal courts must possess some federalcommon-law-making authority before undertaking to craft
it.” 

 And grants of jurisdiction are not wholesale
invitations to create judge-made law. As the Court has said,
“[t]he vesting of jurisdiction in the federal courts does not in
and of itself give rise to authority to formulate federal
common law[.]” Tex. Indus., Inc. v. Radcliff Materials, Inc.,

, 640–41 (1981).
    Why does this history matter? It matters because, when
trying to understand the scope of an old law like the ATS,
history always matters. And given the ATS’s history, we
should acknowledge we don’t have a free hand to create new
causes of action. Indeed, the Supreme Court has outlined



conferred on the federal courts jurisdiction over ‘all suits . . . in equity.’”
Grupo Mexicano de Desarrollo S.A. v. All. Bond Fund, Inc., 

 (1999) (simplified). As a purely jurisdictional grant, the
substantive law had to come from somewhere else. For that, according
to the Court, federal courts are to look to “the principles of the system of
judicial remedies which had been devised and was being administered
by the English Court of Chancery at the time of the separation of the two
countries.” Atlas Life Ins. Co. v. W. I. S., Inc., 

 (1939);
see generally Owen W. Gallogly, Equity’s Constitutional Source, 

 (2023). So looking to the substantive law at the time of
enactment is nothing new.
22                 DOE I V. CISCO SYSTEMS, INC.


several important limiting principles for recognizing new
ATS liability.
     I turn to those next.
                               B.
    In a quartet of cases, the Supreme Court has continually
rejected attempts to expand the list of causes of action
available under the ATS.
    First, consider Sosa. After our own court attempted to
expand the ATS beyond “the handful of international law
cum common law claims understood in 1789” by
recognizing a cause of action for arbitrary arrest and
detention, Sosa, 

, the Court intervened. After
meticulously surveying the history of the ATS and the
“ambient law of [its] era,” 

, the Court warned that
federal courts must exercise “great caution in adapting the
law of nations to private rights,” 

. Reflecting this
caution, the Court established a two-part test before federal
courts may recognize a new cause of action under the ATS.
Because the ATS was meant to “provide a cause of action
for [a] modest number of international law violations,” 

, any new cause of action (1) must stem from violating
“a norm that is specific, universal, and obligatory” under
international law, and (2) must be a proper exercise of
judicial discretion, 

 at 732–33.
    The first step must be undertaken with reference to the
ATS’s historical understanding. Before recognizing “any
claim based on the present-day law of nations,” the Court
said that federal courts must ensure the claim “rest[s] on a
norm of international character accepted by the civilized
world and defined with a specificity comparable to the
features of the 18th-century paradigms we have recognized.”
                 DOE I V. CISCO SYSTEMS, INC.              23

; see also 

 (“[F]ederal courts should not
recognize private claims under federal common law for
violations of any international law norm with less definite
content and acceptance among civilized nations than the
historical paradigms familiar when § 1350 was enacted.”).
     The second step—dealing with discretion—must assess
the “practical consequences” of recognizing a new cause of
action. Id. at 732–33. To that effect, federal courts ought to
consider: “legislative guidance before exercising innovative
authority over substantive law,” id. at 726; “the possible
collateral consequences of making international rules
privately actionable,” id. at 727; the effect of claiming a
“limit on the power of foreign governments over their own
citizens,” id.; and the “risks of adverse foreign policy
consequences,” id. at 728.
    But whatever door the Court opened in Sosa has been
steadily closing, starting with Kiobel v. Royal Dutch
Petroleum Co., 

 (2013). In that case, the Court
rejected ATS jurisdiction over claims against a foreign
corporation when “all the relevant conduct took place
outside the United States.” Kiobel, 

. The
Court applied the presumption against extraterritorial
application—a canon of interpretation for an Act of
Congress—even though the ATS was a jurisdictional statute.

 at 115–16. While the ATS does “not directly regulate
conduct or afford relief,” the Court said that “the principles
underlying the canon of interpretation similarly constrain
courts considering causes of action that may be brought
under the ATS.” 

. Indeed, the Court found it even
more necessary to “constrain courts exercising their power
under the ATS” because of the “danger of unwarranted
judicial interference in the conduct of foreign policy.” 

at 116–17.
24                DOE I V. CISCO SYSTEMS, INC.


     Then comes Jesner. There, the Supreme Court again
affirmed limits on ATS liability. This time, commenting on
the “unresolved” issue of corporate liability, Jesner, 584
U.S. at 251–52, the Court, applying Sosa’s two-step test,
declined to permit foreign corporations to be sued under the
ATS, 

. In doing so, the Court established
the primacy of separation-of-powers concerns in both steps
of the Sosa test. 

. “[S]ome congressional statutes,”
for example, may “bear[] both on the content of the norm
being asserted and the question whether courts should defer
to Congress.” 

. So if there are “sound reasons to
think Congress might doubt the efficacy or necessity of a
damages remedy,” then “courts must refrain from creating
the remedy in order to respect the role of Congress.” 

 (simplified). “[S]eparation-of-powers concerns,” the
Court said, “counsel[ed] against courts creating private
rights of action . . . [particularly] in the context of the ATS.”

 at 264–65. Indeed, the Court even suggested that “Sosa
would preclude courts from ever recognizing any new causes
of action under the ATS.” 

 (emphasis added). But
it left that question for another day.
    Last is Nestlé. Once again, our circuit tried to expand
ATS liability—to domestic corporations for conduct and
injuries occurring almost entirely abroad. Nestlé, 

. Once again, the Court stopped our overreach. 

. While the Court overruled us based on our failure to
require “more domestic conduct than general corporate
activity,” 

, Justice Thomas, joined by Justices
Gorsuch and Kavanaugh, also emphasized that “courts must
refrain from creating a cause of action whenever there is
even a single sound reason to defer to Congress,” 

(plurality). Those Justices reiterated that the “judicial
creation of a cause of action is an extraordinary act that
                  DOE I V. CISCO SYSTEMS, INC.              25


places great stress on the separation of powers.” 

(plurality).
    Under this precedent, several principles emerge. First,
the judicial creation of new causes of action under the ATS
is extraordinarily disfavored if not dead letter. Second, the
historical understanding of the ATS is relevant in assessing
whether a norm gives ground for liability under the statute.
Third, some domestic law, including presumption canons,
govern our interpretation of the ATS’s scope. Fourth,
separation-of-powers concerns almost entirely foreclose the
recognition of new causes of action under the ATS.
   With these principles in mind, I turn to this case.
                              II.
     A group of practitioners of Falun Gong have claimed that
the Chinese government has violated their human rights in
China. They claim that they were subject to the torts of
prolonged arbitrary detention; disappearance; extrajudicial
killing; forced labor; cruel, indecent, or degrading treatment;
and crimes against humanity. But they have not directly
sued China or Chinese officials. Instead, they brought ATS
claims against Cisco Systems, a provider of IT products and
services, for aiding and abetting China’s international-law
violations through its computer-networking hardware and
software. After the district court dismissed the ATS claims,
the panel majority reinstated them under the theory that
aiding-and-abetting liability for any tort violating the law of
nations fell within the ATS’s scope. See Doe, 

. The majority broadly declared that because “aiding and
abetting liability is a norm of customary international law
with sufficient definition and universality,” it can be brought
in a suit under the ATS. 

. Judge Christen
dissented, citing separation-of-powers and foreign-relations
26                 DOE I V. CISCO SYSTEMS, INC.


concerns. 

 at 746–51 (Christen, J., concurring in part and
dissenting in part).
    The panel majority made three critical mistakes. First, it
overlooked the historical scope of accomplice liability by
providing blanket authorization for aiding-and-abetting
liability under the ATS. Second, it upset the separation of
powers by aggrandizing our role in creating law, which the
Constitution grants only to the political branches. And third,
it ignored serious foreign-policy concerns implicated by
expanding aiding-and-abetting liability here. Each of these
errors is sufficient to take this case en banc.
     I turn to each in order.
                                A.
         Historical Aiding-and-Abetting Liability
     To create new liability under the ATS, the Supreme
Court has made clear that we must look to “historical
antecedents.” Sosa, 

. It stated, “federal
courts should not recognize private claims under federal
common law for violations of any international law norm
with less definite content and acceptance among civilized
nations than the historical paradigms familiar when § 1350
was enacted.” Id. (emphasis added). So whatever customary
international law may tell us today, a comparison to
“historical paradigms” within the general common law is
still relevant. On that front, it’s far from clear that aiding- and-abetting liability would attach to all torts claiming
violation of the law of nations. See Khulumani v. Barclay
Nat. Bank Ltd., 

 (2d Cir. 2007) (Korman,
J., dissenting in part) (concluding that historical records “do
not support the extraordinary proposition that Congress
intended the [ATS] to permit jurisdiction to be exercised
                  DOE I V. CISCO SYSTEMS, INC.               27


over claims of aiding-and-abetting without regard to whether
the conduct at issue violated an international law norm”).
    Rather than looking at any historical paradigms, the
panel majority simply endorsed the supposed “consensus”
among modern international-law documents to find
universal accomplice liability. Doe, 

.
Following the Second Circuit, the panel majority considered
the Nuremberg tribunals, unnamed “treaties and
conventions,” U.N. Security Council actions, the
International Criminal Tribunal for Yugoslavia, the
International Criminal Tribunal for Rwanda, and the Rome
Statute of the International Criminal Court. 

 (citing
Khulumani, 

). All these sources, the panel
majority concluded, “recognize some form of accomplice
liability for violations of international law.” 

 On that
basis alone, the panel majority said aiding-and-abetting
liability was “sufficiently definite and universal to be a
viable form of liability under the ATS.” 

 It then ruled
that the liability was available for the seven different torts
alleged by plaintiffs.
    This analysis is incomplete. Whether aiding and abetting
exists as a general matter in international law isn’t the end of
the inquiry. After all, aiding and abetting is not a tort by
itself, but a type of liability that attaches to the commission
of one. See Restatement of Torts § 876(b); E. Trading Co.
v. Refco, Inc., 

 (7th Cir. 2000) (observing
that “there is no tort of aiding and abetting,” rather it is “a
basis for imposing tort liability”). Since aiding and abetting
doesn’t exist in the abstract, the question remains: aiding and
abetting what? See Twitter, Inc. v. Taamneh, 

 (2023) (“The point of aiding and abetting is to impose
liability on those who consciously and culpably participated
in the tort at issue.”) (emphasis added). So it was incorrect
28                DOE I V. CISCO SYSTEMS, INC.


for the panel majority to skip to universal aiding-and-abetting liability without first assessing whether accomplice
liability attached to the underlying conduct at issue for each
tort. Indeed, when considering whether corporate liability
attached, we had remarked that the “analysis proceeds norm-by-norm; there is no categorical rule of corporate immunity
or liability.” Doe I v. Nestlé USA, Inc., 

(9th Cir. 2014) (“Nestlé I”). The same norm-by-norm
analysis should apply for aiding-and-abetting liability.
    At the very least, to determine whether aiding-and-abetting liability applies to every tort in violation of
international law, we should have considered historical
paradigms. While criminal aiding and abetting is an
“ancient . . . doctrine,” civil aiding and abetting has shallow
roots in our law. Cent. Bank, 

. Central Bank
rejected the argument that civil aiding-and-abetting liability
was widespread under domestic law. “[A]t best,” the Court
said, the doctrine was “uncertain in application” in tort law.

 at 181–82 (observing that common-law aiding-and-abetting tort liability was not recognized in Maine,
Pennsylvania, Virgnia, and Montana). Almost derisively,
Central Bank dubbed common-law precedents on civil
aiding and abetting as “isolated acts of adolescents in rural
society.” 

 (simplified). And, unlike with criminal
law, “Congress has not enacted a general civil aiding and
abetting statute—either for suits by the Government . . . or
for suits by private parties.” 

.
    Central Bank thus suggests that civil aiding and abetting
has no storied place in domestic law—whether it be
statutory, common, or general law. So “when Congress
enacts a statute under which a person may sue and recover
damages from a private defendant for the defendant’s
violation of some statutory norm, there is no general
                      DOE I V. CISCO SYSTEMS, INC.                        29


presumption that the plaintiff may also sue aiders and
abettors.” 

    Given the uncertainty of its civil application, we should
have done more before declaring universal aiding-and-abetting liability for all torts under international law. Indeed,
consider the three primary torts accepted at the time of the
ATS’s enactment. While aiding-and-abetting liability was
well recognized for piracy at the time of the First Congress, 3
the record is less definitive for violations of safe conducts 4


3
   For example, in response to a significant “increase[]” in piracy, in
1721, England passed a law expanding liability for “consult[ing],
combin[ing], confederat[ing,] or correspond[ing] with any pirate.”
Piracy Act, 1721, 8 Geo. 1, c. 24 (Eng.). In Colonial America,
defendants were also tried and convicted of aiding and abetting piracy.
See, e.g., Case of John Rose Archer and Others, in John Franklin
Jameson, Privateering and Piracy in the Colonial Period: Illustrative
Documents (1923), at 323–45. And just after passage of the ATS,
Congress also made aiding and abetting piracy a crime. Crimes Act of
1790, ch. 9, § 10, 

, 114 (1790) (defining an accessory to piracy
as “every person who shall . . . knowingly and wittingly aid and assist,
procure, command, counsel or advise any person or persons, to do or
commit any murder or robbery, or other piracy aforesaid, upon the
seas”).
4
  A 1795 opinion from Attorney General William Bradford is often cited
as recognizing accomplice liability for violations of safe conducts. See
Breach of Neutrality, 1 Op. Att’y Gen. 57, 59 (1795). In the opinion,
Bradford was asked whether American citizens who “voluntarily joined,
conducted, aided, and abetted a French fleet in attacking the settlement,
and plundering or destroying the property of British subjects on that
coast” could be subject to criminal penalties. 

. Bradford doubted
that criminal liability attached but suggested that “these acts of hostility
have a remedy by a civil suit in the courts of the United States.” 

.
This short response doesn’t distinguish between principal and
accomplice liability, so we don’t know if he meant to encompass both.
See Khulumani, 504 F.3d at 329–30 (Korman, J., dissenting in part).
30                  DOE I V. CISCO SYSTEMS, INC.


and ambassadorial rights. 5 Without more, this mixed history
signals that aiding-and-abetting liability should not apply
categorically to all torts recognized under customary
international law.
    In response, the panel majority simply dismisses Central
Bank’s analysis because, it claims, the Supreme Court case
does “not govern” the ATS. Doe, 

. Because
the ATS is a “jurisdictional statute,” the panel majority
believed it must rely on its view of “international law, not
statutory text” to decide the “appropriate scope of liability.”

 But that’s off base. As Sosa made clear, any private
claims under the ATS must come from “federal common
law,” 

, so fashioning the scope of liability
must be similarly guided by domestic law—meaning
Central Bank controls. In other words, while international
law may provide the norms at issue, domestic law supplies
the liability for any violation of international norms. See
Kiobel, 569 U.S. at 115–16 (using a canon of interpretation
for federal statutes to limit the ATS’s reach); Nestlé I, 

 (“[I]nternational law defines norms and
determines their scope, but delegates to domestic law the
task of determining the civil consequences of any given
violation of these norms.”) (emphasis added); Louis Henkin,
Foreign Affairs and the United States Constitution 245 (2d

5
     At English common law, committing an offense against an
ambassador—such as by arresting, detaining, or seizing an ambassador,
his servants, or his goods—was an offense punishable under the law of
nations. 4 William Blackstone, Commentaries *70 (1769). The charge
could be brought against “all persons prosecuting, soliciting, or
executing such process.” 

 While Blackstone mentions “soliciting,”
solicitation is different from aiding and abetting and, to my knowledge,
such suits were not commonly filed under an accomplice theory of
liability.
                  DOE I V. CISCO SYSTEMS, INC.              31


ed. 1996) (“International law itself, finally, does not require
any particular reaction to violations of law.”).
    Consistent with the historical caution for recognizing
civil accomplice liability, we should not have accepted so
easily such sweeping aiding-and-abetting liability under the
ATS. To begin, we should have analyzed the liability
available to each underlying tort under international law on
a norm-by-norm basis. And next, we should have looked to
historical paradigms to see whether universal civil aiding- and-abetting liability should be part of our federal common
law. By circumventing this analysis, we skirt the cautionary
roadblocks set by the Supreme Court in creating new causes
of action.
                              B.
                   Separation of Powers
   Even more troublesome here is the affront to the
separation of powers. Not too long ago, I warned:

       Contrary     to    common       belief,    the
       Constitution’s “radical innovation” is not its
       various enumerated rights—as cherished and
       fundamental as they are.           It is the
       Constitution’s design for the separation of
       powers that has become among the “most
       important contributions to human liberty.”
       Having “lived among the ruins of a system of
       intermingled legislative and judicial powers,”
       our Founders sensed the “sharp necessity to
       separate the legislative from the judicial
       power.” The result is the clear division of
       authorities between Congress’s “legislative
32               DOE I V. CISCO SYSTEMS, INC.


       powers”    and    the   Judiciary’s      “judicial
       Power.”

Boule v. Egbert, 

, 373–74 (9th Cir. 2021)
(Bumatay, J., dissenting from the denial of rehearing en
banc) (simplified). So in deciding “which branch of
government may create” new causes of action, the answer
should always be Congress. 

    Notwithstanding Sosa contemplating that federal courts
could recognize a claim based on present-day international
law, the Supreme Court has subsequently signaled that
federal courts should be reluctant to recognize federal
common law tort liability where Congress has not done
so. See Jesner, 

. That’s because “creating a
cause of action is a legislative endeavor.” Egbert v. Boule,

 (2022). At bottom, “[t]o create a new
cause of action is to assign new private rights and
liabilities—a power that is in every meaningful sense an act
of legislation.” 

 (Gorsuch, J., concurring). And as
the Court has instructed in many cases, “[i]f there is a
rational reason to think that the answer is ‘Congress’—as it
will be in most every case”—then we cannot proceed with
the cause of action. 

 (opinion) (simplified).
    While Egbert’s separation-of-powers concern dealt with
Bivens claims, it should sound familiar by now—the Court
has repeated the same concerns for claims under the ATS in
Sosa, Kiobel, Jesner, and Nestlé. See supra Section I.B. The
Court’s message could not be clearer: When creating causes
of action, separation-of-powers concerns predominate and
perhaps even trump all else.
   So, as always, statutory text should control our analysis.
But nothing in the text of the ATS shows congressional
                  DOE I V. CISCO SYSTEMS, INC.              33


intent to advance aiding-and-abetting liability. See 

. There’s no language supporting judicial
creation of new liability. Not a word about which standard
of accomplice liability to use. While this should end our
analysis, we could consider more.
    First, Congress’s inaction speaks volumes. The First
Congress knew exactly how to codify aiding-and-abetting
liability. Just one year after the passage of the ATS, it did
exactly that. In the Crimes Act of 1790, Congress
criminalized accomplice liability for piracy—expanding
liability for any persons who “knowingly and wittingly aid
and assist, procure, command, counsel or advise any person
or persons, to do or commit any murder or robbery, or other
piracy aforesaid, upon the seas[.]” See Crimes Act of 1790,
ch. 9, § 10, 

, 114 (1790). And “[w]hen Congress
knows how to achieve a specific statutory effect, its failure
to do so evinces an intent not to do so.” Fish v. Kobach, 

 (10th Cir. 2016). We should have taken the
hint and left this question for Congress. Instead, we
legislated new liability when Congress did not.
    Second, more recent congressional activity gives more
reason to pause. In 1991, Congress expanded liability under
the ATS but neglected to include accomplice
liability. Congress passed the TVPA, which recognizes a
cause of action for torture. See 

 note. It did
not prohibit aiding and abetting. 

 So Congress opened up
the ATS and included torture. Nothing else. That Congress
went no further should be dispositive. See Nestlé, 

 (Gorsuch, J. concurring) (“The one time Congress
deemed a new ATS action worth having, it created that
action itself[.]”). Justices have also noted the passage of the
Trafficking Victims Protection Reauthorization Act of 2003
and the codification of a private right of action against
34               DOE I V. CISCO SYSTEMS, INC.


perpetrators of human-trafficking violations and slavery as
reason to defer to the political branches in this field. 

 (plurality).
    Third, recall Central Bank. Based on the doctrine’s
uncertain history in tort law and congressional inaction, the
Court dismissed a “general presumption” of reading aiding- and-abetting liability into civil statutes. Cent. Bank, 

. The Court determined that Congress did not
operate with an assumption of such liability in federal civil
statutes. 

 Imposing accomplice liability without any
textual direction would be a “vast expansion of federal law”
without “any precedent,” the Court said. 

. So
Central Bank shows that the statute’s text controls the
availability of accomplice liability. See 

. And
statutory silence means that there is none. 

 Otherwise,
imputing aiding-and-abetting liability would thwart
Congress’s will.
    Thus, courts should not be in the business of creating
causes of action, especially when Congress has spoken (or
not spoken) in this very field. This tells us that universal
accomplice liability is not something that can be slapped
haphazardly onto the ATS. As the Court warned, our
“general practice” must be to look for “legislative guidance”
before innovating in the law. Sosa, 

. Indeed,
“[i]t would be remarkable to take a more aggressive role in
exercising a jurisdiction” than Congress has done over two
centuries. 

 But aggression is exactly what we have
displayed here.
    Despite all these reasons to pause, the panel majority
pressed ahead based on a fundamental misunderstanding of
the judicial role. It looked at whether the political branches
regulated Cisco’s particular activity. See Doe I, 73 F.4th
                  DOE I V. CISCO SYSTEMS, INC.               35


at 723. And seeing a lack of “comprehensive and direct
regulation” of Cisco’s activity by Congress and the
executive branch, the panel majority thought it free to
judicially invent accomplice liability. 

, see also 

 (“Congress and the Executive’s decision not to
regulate or prohibit generally the export of computer
networking software does not conflict with the recognition
that U.S. corporations may be liable . . . for aiding and
abetting violations of international law.”). According to the
panel majority, so long as its creation of aiding-and-abetting
liability doesn’t “displace” or “conflict” with the political
branches’ existing regulatory scheme, then no separation-ofpowers concern exists. 

    But this turns the separation of powers on its head. The
separation of powers means that courts don’t exercise
legislative authority. Period. Just because a judicial
innovation doesn’t affirmatively interfere with the political
branches’ regulatory scheme doesn’t mean it comports with
Article III’s limits. The point is not that political branches
must completely occupy the field of regulation before we
abstain from creating new causes of action. It is that we
don’t belong on the field at all. And in recognizing liability
for violations of international law, that the political branches
entered the field is reason enough for us to exit it. Just
because the political branches left something unregulated—
such as accomplice liability—doesn’t mean we may fill the
void at will. So even assuming the political branches haven’t
comprehensively regulated Cisco’s products and services,
that doesn’t mean we may trade positions with our elected
officials and legislate in the margins. By failing to cabin our
role, we cross into the uncomfortable territory of judicial
lawmaking—an area foreign to our Constitution.
36                DOE I V. CISCO SYSTEMS, INC.


                              C.
                 Foreign Policy Concerns
    Extending aiding-and-abetting liability here raises
foreign policy concerns as obvious as they are serious. The
Court has repeatedly warned of the “inherent” impact on
foreign affairs when creating a cause of action under the
ATS. Jesner, 

. The “principal objective” of
the ATS was to “avoid foreign entanglements by ensuring
the availability of a federal forum where the failure to
provide one might cause another nation to hold the United
States responsible for an injury to a foreign citizen.” 

. So we should be especially reluctant to permit
judicially manufactured causes of action that exacerbate
foreign policy headaches rather than ease them. After all,
the Judiciary has neither the “aptitude, facilities nor
responsibility” to make foreign policy determinations. Chi.
& S. Air Lines, Inc. v. Waterman S.S. Corp., 

 (1948).
     In this case, the plaintiffs seek to hold China and its
government accountable for purported violations of
international law through its alleged aiders and abettors.
Allowing this suit to go forward thus means that a federal
court may participate in declaring that the Chinese
Communist Party and Ministry of Public Security violated
international law in its treatment of Falun Gong
practitioners. As Judge Christen noted in dissent, “[s]uch a
finding could have serious ramifications for Sino-American
relations, fraught as they already are.” Cisco, 

(Christen, J., concurring in part and dissenting in part). So
extending liability here directly risks heightening diplomatic
strife—all the more reason we should have declined to
recognize liability.
                  DOE I V. CISCO SYSTEMS, INC.              37


    The panel majority dismissed these foreign policy
concerns because, it claimed, aiding-and-abetting liability
focuses on the “transgressions of nongovernmental actors”
rather than “the actions of foreign governments themselves.”
Cisco, 

.             But the panel majority
misunderstood the nature of aiding-and-abetting liability.
Such liability is a form of secondary liability, meaning that
someone else must have committed the tortious conduct. See
Restatement of Torts § 876(b). In this case, as plaintiffs
allege, that someone else must be the agents of the Chinese
government. So the only way for this suit to proceed is for
a federal court to adjudicate the responsibility of the Chinese
Communist Party—the ruling party of China—and the
Ministry of Public Security—a government ministry—for
violations of international law. Indeed, “a pronouncement
either way on the legality of other sovereigns’ actions falls
within the realm of delicate foreign policy questions
committed to the political branches.” Spectrum Stores, Inc.
v. Citgo Petrol. Corp., 

 (5th Cir. 2011).
    The panel majority’s answer to the concern for our
country’s relationship with China was largely “so what?”
According to the panel majority, the more important
consideration is that the “international community” might
question our failure to provide a forum to hold accountable
aiders and abettors of international law violations. Cisco, 73
F.4th at 720–21. While that may be true, the Judiciary has
zero competence to make such a decision. This “delicate
[and] complex” question squarely falls within the “domain”
of the political branches and should be left without “judicial
intrusion.” See Chi. & S. Air Lines, 

.
    Two decades ago, in a case similarly alleging
mistreatment of Falun Gong practitioners in China, the U.S.
State Department weighed in on the foreign policy
38                DOE I V. CISCO SYSTEMS, INC.


implications there. See Doe v. Qi, 

(N.D. Cal. 2004). Referring to the ATS and TVPA claims
brought against China, the State Department equivocally
expressed that federal court adjudication of the claims “is not
the best way for the United States to advance the cause for
human rights in China.” 

 at 1271 (quoting the 2002 State
Dep’t Letter). The State Department further urged the
district court in that case to “find[] the suits non-justiciable”
because of the “potentially serious adverse foreign policy
consequences that such litigation can generate.” 

.
In the letter, the State Department cautioned against the
“potential for reciprocal treatment” by Chinese officials
should the suit go forward. 2002 State Dep’t Letter at 8.
While the State Department was concerned about the
treatment of U.S. officials in that case, the implications for
all U.S. citizens abroad in this case are obvious.
    And the panel majority simply drew the wrong
inferences from the fact that the State Department did not
submit a letter here. The panel majority took the State
Department’s lack of affirmative action as a sign that the
“foreign affairs implications here are not comparable” to
other cases involving the Chinese government as a party.
Doe I, 

. But those other cases either involved
Supreme Court cases or cases in which the court requested
the Government’s input. But here, it is “[un]realistic to
expect the Department . . . to monitor all 94 federal district
courts for any ATS litigation raising foreign policy
concerns.” Cisco, 

 (Christen, J., concurring
in part and dissenting in part). What’s more, the panel
majority expressly declined to request the State
Department’s views here. Why? It’s rather baffling. That
should be the minimum when considering creating a new
cause of action with serious foreign policy consequences.
                  DOE I V. CISCO SYSTEMS, INC.              39


    Given all this, it is no wonder that the Government has
long opposed the recognition of aiding and abetting liability
under the ATS. See Brief for United States of America as
Amici Curiae Supporting Petitioners, Nestlé, 

(2020) (Nos. 19-416, 19-453), 23 (“This Court should
conclude—as the government has long argued—that aiding
and abetting liability constitutes an improper expansion of
judicial authority to fashion federal common law under the
ATS.” (simplified)). In particular, the Government was
concerned that “[a]iding-and-abetting claims provide
plaintiffs with a means for evading the limitations of
sovereign immunity and challenging the policies and
conduct of foreign states and officials.” Id. at 25.
    All this provides compelling reason to exercise our
discretion against aiding-and-abetting liability here.
                              III.
     In one single decision, our circuit dismantles the crucial
barrier that separates the three branches of government. We
first undermine the foundation of that barrier by expanding
the ATS beyond its scope and disregarding the clear
action—and inaction—of the Legislature. Then we weaken
this wall by ignoring the Executive’s singular role over
foreign-policy decisions—which we stress by permitting
greater tensions with a world superpower. And finally, we
diminish the respect for the Judiciary by aggrandizing our
role. So just like that, the separation of powers wall comes
tumbling down.
    For these reasons, I respectfully dissent from the denial
of rehearing en banc.