Cisco’s Real Stakes: Digitally Aiding and Abetting
April 16, 2026
This post is cross-published at Just Security.
On April 28, 2026, the U.S. Supreme Court will hear Cisco Systems v. Doe I et al. (Cisco), which asks whether a private U.S. company can ever be sued under the Alien Tort Statute (ATS)—and its CEO sued under the Torture Victim Protection Act (TVPA) (1992)—for aiding and abetting torture and other gross human rights violations. Practitioners of the spiritual movement known as “Falun Gong” and their families have alleged that the U.S. company Cisco Systems Inc. (Cisco) aided and abetted the Chinese Communist Party in subjecting them to torture, detention, extrajudicial execution, and other gross human rights violations. The legal issue before the Court is whether the facts, viewed in the light most favorable to the plaintiffs-respondents, give rise to a cognizable legal action against defendants (who are the petitioners), sufficient to survive a motion to dismiss and allow discovery. The Ninth Circuit answered that question yes, and at the Trump Administration’s urging, the Court granted review.
The case’s long history has now been extensively briefed by the parties, Trump’s Solicitor General (SG), and many amici supporting each side. But as too often happens, this flood of briefing obscures as much as it illuminates. This article explains first, why existing precedent makes Cisco a clear case for affirmance (or dismissal of certiorari as improvidently granted); second, why the Court has no principled reason to overrule settled precedent and thereby immunize U.S. corporate defendants who actively aid and abet mass governmental surveillance that leads to gross human rights abuses; and third, what Cisco portends: the future path of corporate aiding and abetting of mass atrocities through digital means.
The Straightforward Case for Affirmance or Dismissal
On its face, Cisco calls for straightforward application of the Supreme Court’s decision twenty-two years ago in Sosa v. Alvarez-Machain. Sosa interpreted the ATS, first enacted in 1789, which states that federal courts shall have original jurisdiction over a tort “committed in violation of the law of nations or of a treaty of the United States.” Writing for a six-justice majority, Justice Souter held that aliens may sue in the United States under the ATS for claims “based on the present-day law of nations,” so long as those claims “rest on a norm of international character” as “specific, universal, and obligatory” as the traditional causes of action with which the First Congress was familiar. As the amicus brief for professors of legal history filed in the Cisco case (Historians’ Brief) explains, the First Congress was certainly familiar with aiding and abetting, both because it understood the “law of nations” to encompass liability of accessories to crimes like piracy, and because the common law at that time included the “tort” of aiding and abetting.
As demonstrated by the Ninth Circuit, Respondents’ brief, and an amicus brief by international law scholars (including myself), Sosa’s modern-day test is clearly met when an ATS defendant is sued for aiding and abetting such gross violations of international law as torture, extrajudicial killing, and religious persecution. As the amicus brief for U.S. ambassadors and prosecutors reviews, the prohibition on aiding and abetting human rights abuses has been a generally accepted norm of international law since Nuremberg. More generally, aiding-and-abetting (or “accessorial”) liability for violations of international law has been extensively litigated before international tribunals and now ranks among the most firmly established norms of international law. Because aiding and abetting a jus cogens violation violates a generally accepted and specifically defined norm of customary international law just as much as direct perpetration of that act, it belongs in the narrow class of federal common law “torts in violation of the law of nations” that are actionable under the ATS after Sosa. Indeed, the original public meaning of the ATS was that aliens could invoke the statute both to sue pirates directly, as well as those U.S. aiders and abettors who engaged in “furnishing,” “consulting, combining, [or] confederating. . .” with pirates (Blackstone, Commentaries). Holding a U.S. corporation liable in a federal common law action for its domestic collaboration with gross abusers would thus be consistent with the ATS’ common law origins.
But Sosa requires the Court to take a second step as well: to determine whether allowing this case to proceed under the ATS is a “proper exercise of the judicial power.” Here, Cisco, the SG, and their amici warn that hearing such cases would trigger a litany of foreign policy horribles. (12/2025 Brief of SG at 11-12; Petitioners’ Brief at 34-36). But having heard such alarmism for decades—including as the U.S. State Department Legal Adviser and the Assistant Secretary for Human Rights—I find such claims vastly overblown. These claims repeat with little evidence a longstanding tendency to confuse myth and reality regarding the foreign policy interference caused by such corporate responsibility litigation. During three tours in the State Department, I have seen the United States promote aiding-and-abetting liability, as part of consistent U.S. foreign policy, in order to hold perpetrators of international crimes accountable in such far-flung places as the former Yugoslavia, Rwanda, Sierra Leone, Cambodia, Lebanon, Kosovo, and before the International Criminal Court. Cisco and their allies warn that allowing ATS cases to proceed would hurt U.S. commercial activity, undermine U.S. companies, chill foreign investment, and hinder American competitiveness abroad. But petitioners’ brief cites only four claimed examples of foreign policy harm over twenty-two years (Petitioners’ Brief at 35). Not just were all overstated, in one, Doe v. Exxon Mobil Corp. (D.C. Cir. 2011), I filed an expert witness affidavit (nn. 168-69) twenty-five years ago explaining in detail why that claim of harm had no basis.
Exemplifying this overstatement, Cisco and its amici argue that aiding-and-abetting liability will create foreign policy problems, because it requires finding that the Chinese government violated international law (Petitioners’ Brief at 4). But the Chinese government is not a defendant in Cisco. In W.S. Kirkpatrick & Co. v. Environmental Tectonics Corp., International (1990), Justice Scalia dispatched a similar claim for a unanimous Court, noting that “[r]egardless of what the court’s factual findings may suggest as to the legality of [the challenged action under local law], its legality is simply not a question to be decided in the present suit….” Even if it were, the U.S. government has been calling China’s repression of Falun Gong illegal for more than twenty-five years. (See official statements I made more than a quarter-century ago as Assistant Secretary of State for Human Rights). As Respondents note, “[a]dministrations of both parties have condemned the persecution at the heart of this case [and] the United States raises no foreign-policy concerns specific to this litigation.” (Respondents’ Brief at 1) Indeed, the bipartisan U.S. Commission on International Religious Freedom and many NGOs (here, here, here, and here) have long annually tracked the oppressive Chinese government persecution against Falun Gong.
After Sosa, the Court addressed these foreign policy concerns by systematically narrowing the ATS’s territorial reach, holding in Kiobel v. Royal Dutch Petroleum (2013) that corporations may be held liable under the ATS only for extraterritorial actions that “touch and concern” the United States. In Jesner v. Arab Bank (2018), the Court barred the availability of so-called “F-cubed” ATS lawsuits, brought by foreign plaintiffs against foreign corporations for actions occurring in foreign countries. Nestlé v. Doe (2021) then held that even when ATS suits are brought against U.S.-based corporations operating in the United States, allegations of “general corporate activity” or oversight—such as financial decision-making, supervision, or planning—are insufficient to overcome the presumption against extraterritoriality.
But none of these cases forecloses the suit here. Plaintiffs’ second amended complaint alleges far more than just off-the-shelf sales or general corporate activity in the United States. Instead, plaintiffs allege that a U.S. company and its CEO crossed the line from passive supplier to knowing abettor, by custom-designing and tailoring a surveillance system in California for export to China, knowing that the Chinese Government would use that technology to execute its infamous “Golden Shield” surveillance system now widely used in the world’s most surveilled country. A number of credible observers have reported similar allegations—based on leaked internal Cisco documents—that indicate that Cisco knew Golden Shield was being used to persecute Falun Gong practitioners, yet intentionally continued its marketing (here and here). In January 2011, a Boston-based investment firm sold its stake in Cisco, accusing the company of ignoring concerns that its business practices in China might be leading to human rights violations. Congress has repeatedly held hearings on Cisco’s provision of surveillance technology to the Chinese government (here and here), and an amicus brief of bipartisan members of Congress has made similar allegations. None of this is yet proven, and the Cisco defendants would have ample opportunity in discovery and at trial to rebut such allegations on the merits. But the public record so far would seem to render plaintiffs’ complaint plausible enough to surmount the Court’s Twombly/Iqbal standard of pleading and trigger discovery.
The case for discovery becomes even stronger upon recognizing that Cisco’s motion to dismiss does not turn solely on the ATS action. An amicus brief for bipartisan members of Congress clarifies that when a nearly unanimous Congress passed the TVPA, it intended to create a statutory cause of action against any individual who “‘subjects’ [another] to torture.” As the Respondents’ brief notes, the ordinary meaning of one who “subjects” —i.e. exposes or renders another liable to a condition—naturally includes someone who aids and abets torture and extrajudicial killing. Thus, plaintiffs have stated an independent, specific, statutory TVPA claim against Cisco’s CEO, upon which relief can be granted whether or not the ATS claim against the company is dismissed. As the legislative history states, the TVPA authorizes “lawsuits against persons who ordered, abetted, or assisted in the torture” (emphasis added). By so doing, Congress intended that the TVPA implement the nation’s international legal obligations, including Article 4.1 of the Convention Against Torture, which extends the universal prohibition to any “act by any person which constitutes complicity or participation in torture” (emphasis added). An amicus brief by UN Special Rapporteurs on Torture reminds us (at 12-15), that torture is a collective enterprise, “emerg[ing] from systems involving multiple participants, including interrogators, intelligence agents, medical personnel, and logistical facilitators . . . . [P]rivate entities may provide the surveillance technology . . . [as] essential links in the causal chain of abuse” even if they never physically participate in the abusive act. Imposing aiding and abetting liability on those who knowingly provide assistance to torture thus fosters deterrence, promotes accountability, and helps ensure that perpetrators cannot outsource torture to private actors.
Against this straightforward case, Cisco and its allies offer four disconnected objections. First, because the international community has chosen to enforce the aiding and abetting norm through international criminal tribunals, they suggest, the United States may not also impose civil liability via its domestic law.(brief of Washington Law Foundation at 6). Yet the ATS jurisprudence reviewed above shows that that U.S. courts have enforced the civil aiding-and-abetting norm for more than two hundred years, even when criminal liability is being adjudicated in other courts. Second, they argue that recognizing aiding and abetting liability under the ATS raises separation of power concerns. (Petitioners’ Brief at 14). In fact, such concerns run the other way: the federal judiciary has a constitutionally prescribed role to play in foreign affairs, and the First Congress plainly intended the courts, through the ATS, to enforce certain core norms of international law. A far greater violation of separation of powers would occur if courts could eviscerate, through interpretation, statutes that Congress enacted expressly to authorize suits against both perpetrators and their collaborators. Third, Cisco and its allies claim that the Court’s cases since Bivens v. Six Unknown Federal Narcotics Agents (1971) militate against finding a cause of action for aiding and abetting (Petitioners’ Brief at 20). Yet as the amicus brief for federal law scholars shows, the ATS and the TVPA—unlike the Bivens precedents limiting causes of action that are judicially implied from the Constitution—are express congressional grants of authority for courts to recognize claims arising from such universally accepted international law norms as aiding and abetting torture and extrajudicial killing. Fourth and finally, Cisco and the SG repeatedly cite the irrelevant case of Central Bank of Denver v. First Interstate Bank of Denver (1994) (Petitioners’ Brief at 28), which involved neither statute at issue here. Central Bank declined to extend an implied private right of action under the 1934 Securities Exchange Act to aiding and abetting liability. But no U.S. court has ever accepted the argument that Central Bank categorically precludes aiding-and-abetting claims under either the ATS or the TVPA. As the Ninth Circuit found, aiding and abetting is actionable under the ATS because it is a tort in violation of international, not domestic, law. And Central Bank holds only that each statute must be evaluated on its own terms. Here, the deliberately broader terms of the TVPA (note 2(b)) reach anyone who “subjects [another]… to” torture, thereby signaling Congress’ intent to reach beyond immediate perpetrators of torture to collaborators who knowingly act in a way that causes torture and extrajudicial killing.
Even as the U.S. Supreme Court has sought to limit ATS suits against corporate defendants, our democratic allies have moved in the opposite direction, recognizing the availability of corporate liability for human rights violations against parent companies domiciled in those countries. In its 2020 decision in Nevsun Resources Ltd. v. Araya, the Supreme Court of Canada (Abella, J.) held that international human rights norms may impose domestic corporate tort liability on Canadian corporations for adding and abetting gross human abuses abroad. By so holding, Canada followed the 2019 U.K. Supreme Court ruling in Vedanta Resources PLC v. Lungowe, which allowed claims to go forward against a U.K. corporation and its Zambian subsidiary seeking damages for toxic emissions from a mine in Zambia. A Dutch district court held in Kiobel v. Shell (2017) that it had jurisdiction over claims by widows of activists executed in Nigeria who alleged that Dutch and U.K. parent companies and their Nigerian subsidiary were accessories to unlawful arrests, detentions, and executions. Affirmance in Cisco would align U.S. law with these parallel precedents.
Finally, Cisco and its allies claim that letting this aiding-and-abetting case go forward will open the floodgates of new cases. But this is just a variant of the “foreign policy horribles” claim described above. As I have long argued (andRespondents’ Brief at 34-35 confirms), the “floodgates” argument is an unsubstantiated myth that corporate defendants have asserted for decades. Given that every court of appeals that has considered the issue of aiding and abetting has found that the ATS permits such claims (D.C., First, Second, Fourth, Ninth and Eleventh), why haven’t the floodgates already opened? U.S. corporations and their officers have faced the threat of ATS accessorial liability for decades, yet there is no credible evidence that such risks have harmed foreign direct investment or corporate plans to do business in developing countries. An Oxfam amicus brief, signed by several Nobel prizewinners in economics (Nobel Prize Winners Brief at 8-36), debunks the claimed detrimental economic impacts from allowing ATS cases for aiding and abetting. The Nobel prizewinners clarify that recognizing tort liability for aiding and abetting creates appropriate incentives that enhance economic efficiencies by empowering victims—who have the greatest incentive to enforce compliance—while encouraging compliance by those corporate actors best able to police their own conduct. As with the Foreign Corrupt Practices Act, continued U.S. accountability for aiding and abetting would level the playing field for U.S. corporations and promote a “race to the top” promoting better corporate human rights practices. Given that authoritarian foreign governments use systematic human rights violations to suppress domestic wages, the Nobel prizewinners argue that in the long run, imposing aiding-and-abetting liability on U.S. artificial intelligence (AI) and surveillance technology companies will promote better human rights conditions abroad, foster fair trade, and protect U.S. jobs (Nobel Prize Winners Brief at 31-36).
Taken together, all of these considerations should have led the Court to deny certiorari despite the Trump Administration’s suggestion. As the SG acknowledged, five circuits have ruled on aiding and abetting, without creating a circuit split on the core legal issue. (Brief of SG at 2). All parties and the U.S. government condemn Chinese government repression of Falun Gong. The case calls for straightforward interpretation of two extensively litigated statutes, the ATS and the TVPA. The Ninth Circuit correctly applied Sosa’s two-decades-old controlling precedent, which has clarified the legal standard and contained the growth of subsequent litigation. Foreign courts have agreed that aiding-and-abetting liability can be made available, without unleashing a new wave of lawsuits. Non-meritorious cases can be carved out with procedural scalpels—such as personal jurisdiction, forum non conveniens, and misjoinder—rather than applying a premature jurisdictional meat axe to bar all aiding-and-abetting claims forever. On its face, a Court that professes to be committed to textualism, originalism, strict construction, and the passive virtues should either affirm or dismiss cert as improvidently granted.
Applying Stare Decisis
Nor, despite Cisco’s suggestion (Petitioners’ brief at 17-18), does this case provide a vehicle for overruling Sosa. Cisco and its allies urge the view that the ATS is frozen in time, with the list of cognizable “torts in violation of the law of nations” restricted to the three original violations recognized in 1789 (piracy, attacks on diplomats, and violations of safe conduct). But this same construction of the statute was rejected by six justices in Sosa, and the words of the statute have not changed. Under this frozen view, a traveler denied safe conduct could recover under the ATS, but a victim of modern genocide perpetrated from United States territory could not. Even if the Sosa Court had not already rejected this position, the Historians’ brief (at 14-15, 26-31) confirms that the First Congress intended for the ATS to evolve along with the law of nations, not stay fixed only to the law-of-nations violations that existed as of 1789.
Calling it “a jolt to the legal system when you overrule a precedent,” Chief Justice Roberts has cautioned that “[t]he principle of stare decisis has ‘special force’ ‘in respect to statutory interpretation’ because ‘Congress remains free to alter what we have done.’” As an academic, then-Professor Amy Coney Barrett explained the Supreme Court’s two reasons for affording its precedents interpreting statutes (like Sosa) special protection from overruling. First, when as here, congressional silence follows the Supreme Court’s interpretation of a statute, that signals “a refusal to veer from an interpretation that Congress has effectively approved.” Second, if dissenting Justices disagree with a standing Supreme Court precedent, they should refuse to revisit that statutory interpretation to encourage Congress to amend the statute through the democratic process, rather than resolving the interpretive dispute via judicial process.
Here, Cisco and its amici offer no good reason to overcome this super-strong presumption of statutory stare decisis. The statutory standard has not changed since Sosa. Nor, despite some Justices’ claims (here at 1406-07) is there any basis for upsetting settled law by claiming that the Erie doctrine somehow precludes the federal courts from making federal common law to enforce international human rights law. Justice Souter’s majority opinion in Sosa long ago resolved this old academic debate, stating: “Erie did not in terms bar any judicial recognition of new substantive rules, no matter what the circumstances, and post-Erie understanding has identified limited enclaves in which federal courts may derive some substantive law in a common law way.” Despite changes in the Court’s membership, claims under customary international law “arise under” federal law for purposes of Article III of the Constitution. Since Sosa, the Court has twice decided ATS cases between two aliens on the merits (Jesner and Kiobel), thus confirming that federal courts may exercise Article III jurisdiction over torts in violation of international law as federal questions.
Cisco and its allies thus offer no basis for overruling Sosa, which articulated a stable, well-functioning legal standard that serves the First Congress’s purpose. Sosa has not undermined, but instead reaffirmed, important legislative objectives, such as providing a civil remedy to victims of torture. The unanimous post-Sosa circuit rulings authorizing aiding-and-abetting liability have not created unworkable confusion. Like-minded democratic allies also enforce that norm through civil liability in their domestic courts, even while it is also being enforced by international criminal tribunals. Despite Cisco’s warning, domestic judicial enforcement of this body of law has not unleashed the floodgates of litigation. Judicial enforcement of the ATS and TVPA is not only consistent with the will of Congress, but also with the textualist mode of statutory interpretation endorsed by Justices Gorsuch and Barrett. No judicial or foreign policy “emergency” suddenly warrants the activist step of abruptly reversing the Court’s settled 6-3, decades-old statutory interpretation precedent.
The Real Issue in Cisco
The legal issue in Cisco arises from the disparity between factual allegations in the second amended complaint—which on a motion to dismiss must be accepted as true—and Cisco’s denial of those allegations, which can only be resolved through discovery. Cisco claims (Petitioners’ Brief at 9-10) that “it sold only off-the-shelf networking equipment to the Chinese government, in compliance with United States export regulations, and that it did not customize that equipment.” If proven, such conduct probably would not arise to an actionable claim of “aiding and abetting” under the applicable international law standard, which the amicus brief for international law scholars (at 15) argues would require “(1) assistance that has a substantial effect on the perpetration of the principal’s offense (the actus reus); and (2) knowledge that the acts assist the commission of the offense (the mens rea).”
But plaintiffs allege (Respondents’ Brief at 3-7) that Cisco went further: that defendants were fully aware that Chinese security services would use that system to surveil, track, apprehend, and interrogate Falun Gong practitioners to execute arbitrary detention, torture, extrajudicial killings, and forced religious conversion. As evidence of this knowledge, Respondents allege that Cisco prepared marketing materials and internal reports, such as PowerPoints, brochures, and other files demonstrating that Cisco knew about the Chinese government’s persecution and intentionally tailored its technology to aid the persecution of Falun Gong practitioners (see Second Amended complaint at the District Court level, paras. 58-74, 183-195). If proven, such detailed allegations would satisfy the requisite “substantial effect/knowledge” test for aiding and abetting and exceed the level of “general corporate activity in the United States” called for by Nestlé.
To affirm or dismiss cert would not foreclose whatever meritorious defenses Cisco may have; it would only leave open a door leading to more litigation. In a case that may follow Cisco to the Court, Al Shimari v. CACI Premier Technology, Inc., after 17.5 years and five appeals, the Fourth Circuit affirmed a $42 million jury verdict holding a U.S. company liable under the ATS for conspiracy to commit torture and cruel, inhuman, and degrading treatment. The jury found that a U.S. company had employed interrogators who worked with U.S. military officials to help abuse detainees at Abu Ghraib prison in Iraq, grossly violating detainees’ human rights by subjecting them to threats, sexual assault, forced nudity, dog attacks, and prolonged stress positions. What Al Shimari shows is that if a meritorious ATS case can clear all procedural hurdles and make it all the way to a jury verdict, the corporations will have more than ample opportunity to defend themselves on the merits.
So Cisco presents an asymmetric choice: between trusting the lower courts to sort out a complex, evolving issue versus premature jurisdictional overkill. If the Court rushes to intervene at this jurisdictional stage—and holds that civil liability is never available against U.S. companies that knowingly provide technology with the substantial effect of aiding and abetting repression achieved through mass surveillance—the only protection left will be corporate self-policing.
In the end, Cisco is not—as petitioners would have the Court believe—just about protecting U.S. corporations from frivolous lawsuits. It really asks whether acknowledged victims of an American company’s knowing aiding and abetting may ever invoke settled precedent to pursue federal recovery for intentional corporate actions taken within the territorial United States that actively assist an intrusive government to commit gross human rights violations. Cisco’s counsel must answer the question: “If victims of torture and other human rights abuse could prove it to a jury, why shouldn’t they be allowed under Sosa to sue an American corporation that knowingly aided and abetted torture and other gross human rights violations abroad?”
Beyond Cisco
Cisco’s implications go far beyond its immediate facts. A momentous issue that all societies now face is the frightening prospect that new technologies will enable highly-effective mass surveillance—and with it, more effective governmental repression. Discouraging technology companies from actively enabling such repression will be key to the future of human liberties, not just abroad, but here in the U.S. as well.
In this regard, the most telling amicus brief in Cisco comes from the Electronic Frontier Foundation (EFF), which chronicles how U.S. companies have allegedly provided the tools and technologies to enable the global human rights abuses. Both the first Trump administration and the Biden administration endorsed efforts to stop the use of digital technologies and surveillance to violate human rights. Even beyond Cisco, the EFF brief points out, other U.S. technology companies—including Yahoo!, IBM, Oracle, Dell, HP, ArcGIS, and Microsoft—allegedly pitched their technology as tools for Chinese police to build China’s surveillance state (EFF Brief at 12-14). Beyond China, EFF charges, U.S. technology companies have contributed to human rights abuses by such diverse regimes as the South African apartheid government and the Belarusian, Egyptian, Emirati, Israeli, Saudi, Syrian, and Tunisian governments (EFF Brief at 15-22). As repressive regimes around the world increasingly apply digital surveillance to unlawfully arrest, detain, torture, disappear, and summarily execute victims, the Justices should wonder whether adopting Cisco’s bright-line exemption of even active aiders and abettors would turn the United States into a safe harbor for digital abusers.
No less urgently, the Justices should ask whether affording such immunity in Cisco would implicitly signal to U.S. companies that they could not be held liable for assisting the U.S. government in committing gross violations. Such violations could include constructing a mass surveillance system that may be used to surveil, persecute, retaliate against, and punish real and perceived opponents and constructing systems that autonomously commit war crimes. Three high-profile Trump Administration episodes suggest that this is no longer a hypothetical concern:
Mass Surveillance in ICE Raids: As the nation’s most heavily funded law enforcement agency, the U.S. Immigration and Customs Enforcement (ICE) has substantially expanded its deployment of advanced surveillance technologies to use facial recognition software like Mobile Fortify, biometric trackers, mobile phone location databases, and social media monitoring to promote the Trump Administration’s initiative on mass deportation. According to a complaint recently filed by Illinois and Chicago, ICE has utilized Mobile Fortify in the field more than 100,000 times, constituting a major expansion in the technology’s use (Complaint para. 109). This technological expansion could just as easily be deployed to promote mass surveillance on campuses and at nationwide protests, raising grave Fourth Amendment concerns regarding individual privacy.
Government Overriding AI Companies’ Restrictions on Mass Surveillance: As the Trump Administration’s current legal battle against the AI company Anthropic shows—and as a bipartisan amicus brief by 16 former senior national security officials (co-authored by myself) argues—Anthropic’s current litigation asks whether the United States may punish an American AI company as a “supply chain risk” for refusing to allow the government to use its AI, inter alia, for such unlawful ends as committing mass atrocities abroad. Should the Administration ultimately prevail in Anthropic—where a California district judge recently granted, and a D.C. Circuit panel denied, a stay to lift the “supply chain risk” designation pending expedited appeal—government mass surveillance by conscripting commercial AI software could proliferate.
Digitally Aiding and Abetting War Crimes: Finally, as I have elsewhere discussed (at 16-19), AI presents the growing possibility that autonomous systems may be used in the future to commit war crimes abroad. The Department of Defense has demanded—as a condition of continuing as a Department contractor—that Anthropic must allow its technology to be used not just for mass surveillance, but to develop fully autonomous lethal weapons. After Anthropic refused to abandon its core principles against such uses, Defense Secretary Pete Hegseth not only canceled the Department’s contract with Anthropic, but labeled it as what the D.C. Circuit panel called “an unwanted vendor,” because the company “has now conclusively barred uses that the Department recently deemed essential.” But what law authorizes Hegseth—a Cabinet member sworn to uphold the Constitution and laws of the United States—to deem it “essential” that U.S. military software strip out private protections against executing lethal autonomous strikes?
As of November 2024, 57 countries, led by the United States, had endorsed the Political Declaration on Responsible Military Use of Artificial Intelligence and Autonomy. That Declaration states that “[m]ilitary use of AI must be in compliance with applicable international law [and] use of AI in armed conflict must be in accord with States’ obligations under international humanitarian law, including its fundamental principles.” More than 100 international law experts recently declared that the first weeks of the Trump Administration’s military campaign in Iran exhibited apparent violations of the Geneva Conventions. Even as Hegseth was disparaging Anthropic as an unwanted vendor, his Department was reportedlyusing its AI program Claude for intelligence assessments, target identification, and simulating battle scenarios in launching its military campaign in Iran. So yet another question the Justices should ask Cisco’s and the U.S. government’s lawyers is: “If a foreign or U.S. government refuses to obey the Geneva Conventions while enlisting AI technologies in military conflict, does a victim have no civil remedy against an American company that aids and abets the government’s commission of war crimes?”
Conclusion
The Court should not have granted cert in Cisco, because the decision below was correct, and the circuits were not split on the key issues. Affirmance or dismissal of cert would simply allow the lower courts to clarify the nuances of digital aiding and abetting in future cases, depending on the gravity of the atrocities alleged and the technology being deployed.
That outcome would promote corporate social responsibility. Responsible technology companies could mitigate the risk of aiding and abetting lawsuits by, for example, implementing more robust human rights due diligence frameworks that align with the UN Guiding Principles on Business and Human Rights, requiring risk assessments before signing contracts with governments, and (like Anthropic) incorporating human rights considerations into their development of surveillance technologies. But as Anthropic’s litigation shows, governments who seek to enlist private technology to promote mass surveillance—including the U.S. government under this Administration—may seek coercively to override internal corporate ethics, gravely threatening individual privacy and security. Aiding-and-abetting liability is a necessary bulwark to prevent that from happening.