AMNESTY’S GENOCIDE EXCEPTION
How Amnesty International applied a historic legal label only to Israel & failed to live up to its own standards
With South Africa’s genocide case taking on a slow pace before the International Court of Justice (ICJ), it is worth revisiting the genocide report that for many cemented claims that “the experts” have concluded that Israel is guilty of the ultimate crime in international law. That is, Amnesty International’s most downloaded report in history: a 296-page document concluding that Israel had committed, and was continuing to commit, genocide against Palestinians in Gaza. Amnesty’s Secretary General, Agnès Callamard, called the finding “unequivocal.”
Amnesty’s report was unprecedented in a very specific and legally significant way: it was the first time in its history that the organization independently, and without any prior finding from an international court, concluded that a genocide had taken place.
Its uniqueness warrants scrutiny. And more so, the pattern surrounding it demands a reckoning. Amnesty’s methodology is that for Israel, the burden of proof for the crime of genocide must be lowered in order to convict. But for all other countries and atrocities, we must maintain the highest evidentiary burden. It would be as if someone argued that “proof beyond a reasonable doubt” is too burdensome for the prosecution and therefore, we must lower it to send more people to jail, but only if they’re from a certain racial group. Pointing out the hypocrisy is not ‘whataboutery’; it is about whether we can trust that laws and standards will be fairly applied. Human rights work is important, and it is undermined when legal standards are selectively changed or selectively applied for what looks like purely partisan ends.
The Standard Amnesty Invoked
Under the 1948 Genocide Convention, genocide requires proof that acts were committed with dolus specialis, that is the specific intent to destroy a national, ethnic, racial, or religious group in whole or in part. The International Court of Justice (ICJ) has held that this intent must be the “only reasonable inference” from the evidence. This is a high bar.
Taking Amnesty at its word and applying its own framework consistently across comparable cases reveals some glaring inconsistencies.
Amnesty Lowered the Bar — but only for Israel
Amnesty did not simply apply the ICJ standard to Israel. It first attacked the stringency of that standard as “overly cramped,” and then applied a revised, lower standard exclusively to Israel.
The ICJ’s established test, articulated in Bosnia v. Serbia (2007) (para 373) (and reaffirmed in Croatia v. Serbia (2015), paras. 145-148), requires that to infer genocidal intent from a pattern of conduct, it must be “the only reasonable inference which can be drawn from the pattern of conduct.” All other plausible explanations (military necessity, indifference, command failure, even other atrocities such as ethnic cleansing) must be exhausted before inferring the specific intent to destroy a group as such.
Amnesty’s report resisted this standard. On page 101 of the 296-page document, Amnesty characterized the ICJ’s test as “an overly cramped interpretation of international jurisprudence,” one that “would effectively preclude a finding of genocide in the context of an armed conflict.” In its place, Amnesty substituted a “holistic” assessment under which genocidal intent need not be the only reasonable inference (“ORI”); it can coexist alongside military goals. Callamard explicitly stated that “genocidal intent can co-exist alongside military goals and does not need to be Israel’s sole intent.” Nevertheless, Amnesty maintained that genocidal intent was still the “only reasonable inference” in Gaza. But it reached that inference without exhausting the competing military-necessity explanations the ORI test requires it to rule out. Amnesty’s approach “effectively applies an alternative test not based on the established jurisprudence of the ICJ.”
The report approvingly cited to the intervening states in The Gambia v. Myanmar case, who argued that the ICJ’s genocide jurisprudence is too stringent. But, as we will see later, Amnesty itself refuses to apply this “less stringent” standard to Myanmar’s atrocities against the Rohingya. For Amnesty, Myanmar can “get away with it”; Israel must not.
Amnesty has never applied its “holistic” or “cumulative” intent framework to any other situation examined in this article. In El Fasher, Myanmar, Darfur, Ukraine, and Xinjiang, Amnesty’s characterizations have not risen above crimes against humanity, ethnic cleansing, and war crimes. Amnesty never applied the dual-intent analysis it used for Gaza. The lowered evidentiary threshold was introduced specifically for, and applied exclusively to, Israel.
Amnesty’s Conservative Approach to Every Other Atrocity
Amnesty’s innovation for Gaza was not about which enumerated act occurred; it was about how to infer specific intent from a pattern of conduct. That method is not act-specific. Dolus specialis is the common element across every act in Article II, so the holistic inference is, by its own logic, portable to any genocide analysis: whether the underlying act is killing, inflicting destructive conditions, preventing births, or forcibly transferring children. The question, then, is not why Amnesty reached different conclusions about what happened in different places, but why it reserved its more permissive test for intent for one state only.
Amnesty pitched that test as a fix for a specific problem: that “in the context of an armed conflict,” the strict standard would “effectively preclude a finding of genocide,” because military necessity always supplies a competing inference. That rationale maps to the armed-conflict killing cases in El Fasher, Myanmar, Darfur, as well as Ukraine (though there the genocidal action involves abduction of children), which is why Amnesty’s refusal to extend the methodology to any of them, even where a UN body applying the stricter test found genocidal intent, is clear evidence of selectivity.
El Fasher, Sudan (2025). The RSF’s October 2025 offensive produced more than 6,000 documented killings in three days, with the UN noting the true toll was “undoubtedly significantly higher.” Killing was ethnically targeted at Zaghawa and Fur communities. Survivors were asked their ethnicity before being executed. RSF perpetrators explicitly called for the elimination of non-Arab communities.
This is the cleanest of the comparisons, because the two institutions assessed the same conduct against the same legal test and diverged. In February 2026, the UN Independent International Fact-Finding Mission for Sudan found that the takeover bore the “hallmarks of genocide” and, after expressly considering and rejecting alternative explanations, concluded that genocidal intent was the only reasonable inference that could be drawn from the pattern of conduct against the Zaghawa and Fur. That is the very “only reasonable inference” test Amnesty dismissed as too stringent when assessing Israel.
But Amnesty, assessing the same events, did not find genocide either through the stringent standard or its advocated for (in Gaza) less stringent standard. In its contemporaneous November 2025 report on the fall of El Fasher, Amnesty’s Callamard characterized the violence as conduct that “constitutes war crimes and may also constitute other crimes under international law.” Amnesty had not applied the genocide label. And, more to the point, Amnesty did not apply to El Fasher the “holistic” dual-intent analysis it had built for Gaza. (Update, August 2026: Amnesty’s comprehensive El Fasher report, published July 2026 after this article first appeared, again concluded crimes against humanity and ethnic cleansing, with genocide only being “under investigation.” This is still short of the finding a UN body had reached five months earlier under the stricter test.)
Myanmar. In 2017, Myanmar’s military drove approximately 725,000 Rohingya from their homes in weeks through mass killings, systematic sexual violence, and village burning. The UN’s own Fact-Finding Mission concluded that Myanmar’s military leadership “must be investigated and prosecuted for genocide.” Amnesty’s characterization, in its 2018 report We Will Destroy Everything: Military Responsibility for Crimes against Humanity in Rakhine State, was crimes against humanity and ethnic cleansing. Not genocide. On the eight-year anniversary in 2025, Amnesty reiterated that framing, noting that while the UN Fact-Finding Mission had documented evidence of genocide, its own characterization remained crimes against humanity and war crimes.
Darfur. The conflict beginning in 2003 produced approximately 300,000 deaths and 2.7 million displaced. The United States determined that genocide had occurred and the ICC ultimately charged President al-Bashir with genocide.
In 2004–05, Amnesty’s decision to characterize the atrocities as grave abuses and crimes against humanity, rather than genocide, was consistent with the UN’s own legal body: the International Commission of Inquiry on Darfur concluded in January 2005 that the Government of Sudan had not pursued a policy of genocide, while finding crimes against humanity and war crimes potentially no less grave. That contemporaneous conclusion also predates the ICJ’s “only reasonable inference” test, which was not articulated until Bosnia v. Serbia in 2007. What is notable is that in the two decades since, and across a continuous record of Darfur reporting (e.g., Amnesty’s July 2026 report on El Fasher), Amnesty has never issued an affirmative genocide finding on Darfur, even as the surrounding jurisprudence, including its own, changed.
China and the Uyghurs. Amnesty published a 160-page report in 2021 documenting systematic mass imprisonment, torture, and persecution of over one million Uyghurs. Amnesty’s conclusion: crimes against humanity. The U.S. government, by contrast, declared it genocide in January 2021. Other Western parliaments have done the same, such as the UK, Canada, and the Netherlands. Amnesty has not. (Amnesty’s own view re Gaza, that the ICJ bar is “too stringent” cuts against invoking stringency selectively here.)
Russia and Ukraine. Russia’s forcible transfer of Ukrainian children to Russia is one of the enumerated genocidal acts under Article II(e) of the Genocide Convention. Amnesty’s characterization: war crimes and likely crimes against humanity. Not genocide. (This is an armed conflict, and Amnesty’s own “intent can coexist with other goals” logic applies here. But Amnesty declined to use it.)
The October 7 Problem
Amnesty’s differential application runs in both directions. In December 2025, over two years after the attack, Amnesty published its first comprehensive report on Hamas’s October 7 conduct, concluding Hamas committed the “crime against humanity of extermination.” But not genocide.
However, applying Amnesty’s own “instrumental intent” framework, Hamas’s documented ideology and subsequent actions satisfy the standard for genocide at least as readily as the inference Amnesty drew regarding Israel.
Intent & Genocidal Rhetoric. Rhetoric calling for the annihilation, not only of the State of Israel, but the Jews living in Israel, and even outside of it, is legion:
Hamas’s 1988 founding charter explicitly calls on Muslims to kill Jews.
A 2021 conference held in Gaza envisioned a future Palestine in which an indeterminate number of Jews had been killed, but with some Jews remaining in the country as slaves.
A Hamas official, Hamad Al-Regeb, delivered a sermon in April 2023 describing Jews as “filthy, ugly animals like apes and pigs” and praying for their annihilation.
Hamas leader Mohammed Deif’s October 7 call to “kill them wherever you find them” (also documented in a peer-reviewed legal analysis as establishing genocidal intent under the dolus specialis standard).
Senior Hamas official Ghazi Hamad vowed to “repeat the October 7 attack time and again.”
An IDF-released phone recording — played to the UN Security Council — captured a Hamas fighter announcing he had killed Jews with his bare hands. And in intercepted battlefield communications, Hamas commanders urged their fighters to “Slaughter them. End the children of Israel.”
Hateful rhetoric towards Palestinians was key to Amnesty’s accusation of genocide by Israel. But hateful and genocidal rhetoric towards Jews and Israelis, followed by violent action, is absent from Amnesty’s analysis of the crimes of October 7. Beyond that, Amnesty not only refuses to condemn antisemitic and violent rhetoric by Hamas or other anti-Israel groups, it has also opposed requirements imposed by European states that Palestinian civil society organizations condemn Hamas. Amnesty deems such requirements “an attack on the human right to freedom of expression and association.” No such position has been taken regarding counter-extremism requirements directed at Israeli or Jewish organizations.
The Srebrenica Precedent. The ICTY treated the execution of 8,000 military-age Bosnian Muslim men in Srebrenica, together with the forcible transfer of the women, children, and elderly, as the evidence from which the intent to destroy that entire community could be inferred. (See the Krstić Appeals Judgement, IT-98-33-A.). Krstić stands for two propositions relevant here: first, the targeted “part” may be delimited by the perpetrator’s geographic area of operation; second, genocidal intent may be inferred from a pattern of killing combined with related conduct against that part.
Applied to October 7, the inference is carried by the documented intent evidence set out above: the annihilationist rhetoric, the ethnically defined targeting, and the methodical conduct within the communities Hamas physically reached. On October 7, Hamas massacred 378 people at the Nova music festival, and in Kibbutz Be’eri its fighters “roamed freely. . . killing methodically and with cruelty”, murdering roughly 10% of the community. Hamas’s combination of both mass and selective killings, although not at the numerical scale of Srebrenica, alongside the abduction of others, including children, is the kind of conduct the ICTY found probative of intent. Amnesty however, characterized these events as “just” crimes against humanity; it did not undertake the intent inference at all, though it undertook precisely such an inference, on thinner and more ambiguous rhetorical evidence, for Israel.
Ignoring its Own Prior Reports
Amnesty also failed to investigate whether Hamas continued its practice, as documented in a 2015 Amnesty report, of using al-Shifa Hospital as a detention and torture facility. And it ignored U.S. intelligence reports that Hamas and other armed groups used medical facilities for military operations. Nor did Amnesty seriously examine, as it previously had, whether Hamas was continuing to endanger Palestinian civilians by launching offensive operations from residential areas. The UN Secretary General, hardly a pro-Israel figure, condemned Hamas’s use of human shields in November 2023. Amnesty’s refusal to give detailed consideration to Hamas’s use of civilian infrastructure for military purposes is directly relevant to the proportionality and intent analysis at the center of its genocide finding against Israel.
The Personnel Question
The lead researcher on Amnesty’s Israel/OPT portfolio is Budour Hassan. According to her own biography, her entire professional career prior to joining Amnesty in 2022 was at the Jerusalem Legal Aid and Human Rights Center, a Palestinian civil society organization. She has described her advocacy work as being about “solidarity — naming the perpetrator, rejecting silence.”
The researcher identified by NPR as having analyzed Israeli official statements for the genocide report is Kristine Beckerle, Amnesty’s Deputy Regional Director for the Middle East. She lives in Beirut, holds a Yale Law degree, and previously worked at UNRWA, the UN agency for Palestine refugees, whose institutional integrity has been the subject of much scrutiny.
The question is not whether these researchers are competent, one does not have reason to believe they are not, or that they are personally biased. Rather, it is whether Amnesty, in producing the most consequential legal finding in its history, took the ordinary step of insulating that finding from the appearance of partiality. In a courtroom setting, a judge would recuse himself or herself if impartiality might be reasonably questioned such as by having a prior professional affiliation to a party-in-interest. This is done to protect the integrity of the process and its result. Here, the two researchers most closely associated with the finding had prior institutional, Palestinian-specific advocacy ties bearing on the subject, and Amnesty assigned them without any visible safeguard.
The contrast with how Amnesty staffed its comparable reports on China and Ukraine makes the point precisely. The lead author of Amnesty’s 2021 Xinjiang crimes against humanity report was Jonathan Loeb, a Senior Crisis Adviser who previously conducted research on conflict in Sudan. He has no documented background in anti-China activism or Uyghur advocacy. Amnesty’s Ukraine field reporting has been led by Donatella Rovera, an Italian national with over 20 years at Amnesty whose prior missions covered Syria, Libya, Iraq, Yemen, Sudan, and Somalia. No Ukrainian nationality, no Ukrainian advocacy affiliations. And, despite documenting Russian atrocities against the Ukrainian people, Ms. Rovera’s team still managed to condemn Ukraine’s own government of war crimes for the very thing Amnesty is reticent to do in Gaza. This is not to reject all use of regionally embedded researchers, without which, NGOs like Amnesty could not work. However, for a first-ever, precedent-setting genocide determination Amnesty declined to match even the arm’s-length impartiality its own comparable reports reflect.
Punishing Internal Dissent
Amnesty’s own Israel branch rejected the genocide determination. According to reporting in Haaretz, internal correspondence referred to the document as the “genocide report” before research was complete, and staff members accused it of being “motivated by a desire to support a popular narrative among Amnesty International’s target audience.”
Amnesty’s response: it suspended the Israel branch for two years. The branch with direct operational knowledge of the conflict was silenced. The researchers who produced the report faced no equivalent accountability.
The Downstream Effect
South Africa’s February 2025 diplomatic dossier to the UN Security Council, filed in connection with South Africa v. Israel before the ICJ, cited the Amnesty report directly. South Africa formally welcomed the report as affirming its ICJ case. Amnesty’s written evidence to the UK Parliament cited the genocide finding as the legal basis for calling on the government to suspend arms transfers. Amnesty USA used it to lobby for the Block the Bombs Act.
This cascade occurred without the methodological concerns documented above ever being publicly disclosed or adjudicated. A report whose lead researcher had prior ties bearing on the subject, whose institutional predetermination was documented in internal communications, and whose dissenting affiliate was suspended, was incorporated into international legal proceedings and legislative records as authoritative evidence.
The media amplification layer. The institutional cascade was substantially amplified by media figures who made the genocide framing central to their platforms. Amnesty USA’s executive director Paul O’Brien gave his first interview about the genocide report to Mehdi Hasan’s Zeteo, stating the Palestinian people “will not recover from this in our lifetimes as a people, and we’re failing to call it what it is.” Democracy Now! hosted Amnesty researcher Budour Hassan the day after the report’s release with no critical counterpoint on methodology. Media Lens published a widely circulated piece criticizing the BBC for insufficient amplification of Amnesty’s findings. For much of the media, and certainly among Leftist channels, Amnesty’s report is final and authoritative; no counterpoints have been raised about its methodology.
The “genocide” framing has achieved majority or near-majority acceptance in the English-speaking democracies — a shift driven substantially by the institutional cascade documented above. A YouGov poll (UK, June 2025): 45 percent of UK adults believe Israel’s actions amount to genocide. A Quinnipiac poll (US, August 2025): 50 percent of US voters, including 77 percent of Democrats. An Angus Reid poll (Canada, August 2025): 52 percent of Canadians (an 11-point rise in 18 months). All three shifts occurred in a period in which the primary institutional authorities cited (Amnesty, HRW, the UN Commission of Inquiry) were applying the genocide label to Gaza while withholding it from equally or more severe contemporaneous atrocities. That methodological concern has received almost no mainstream examination.
Undermining International Humanitarian Law and Public Trust
Human Rights Watch, it should be noted, exhibits the same pattern: it found “acts of genocide” specific to water deprivation in Gaza while characterizing Darfur, the Uyghur persecution, and the El Fasher massacre as crimes against humanity. HRW’s own World Report 2024 acknowledged that selective application of human rights law “weakens the belief in the universality of human rights.”
The argument made here is not that Israel is innocent of all wrongdoing. But Amnesty International applied the genocide label to Israel under a “holistic” intent framework it has never applied to Myanmar, China, Russia, or Sudan. And even when UN investigators found genocidal intent in El Fasher, applying the strict “only reasonable inference” test that Amnesty dismissed as “overly cramped” when judging Israel, Amnesty did not budge from characterizing the same conduct by the less severe term “war crimes.” Amnesty has identified no facts to dispute the UN’s El Fasher findings, so the divergence between them cannot be explained by the facts but by the label each institution was willing to apply – and why. Surely, if even under the strictest standard, genocide can be found in El Fasher, then Amnesty’s less strict standard can, and should, find the same. Yet Amnesty remains extremely conservative in its approach to that conflict.
The method Amnesty built for Gaza was a claim about inferring intent, not about any particular act. That it was never carried anywhere else is revealing. This refusal is starkest in the cases the method was expressly designed for: armed-conflict mass killings (Myanmar, Darfur, El Fasher). And it is nearly as stark for Russia, where the same coexisting-intent logic maps onto the forcible transfer of children. The difference here is not explained by the evidence, by timing, or by urgency. The variable that best predicts when Amnesty reaches for the genocide label, and the more permissive intent test that unlocks it, is the identity of the perpetrator and the identity of the victim group.
That is not how international law is supposed to work. And Amnesty International knows it because it once said so itself: “Countries applied human rights law on a case-by-case basis in a staggering show of blatant hypocrisy and double standards. . . . It’s unconscionable and undermines the entire fabric of universal human rights.” Callamard was describing governments. The evidence suggests she was also describing her own organization.