How the Geneva Conventions actually work in practice — versus in theory

On August 12, 2024, ICRC President Mirjana Spoljaric Egger issued a statement marking the 75th anniversary of the Geneva Conventions. She called international humanitarian law “a universal framework” and urged states to recommit to its principles. “In a divided world,” she said, “the Geneva Conventions and international humanitarian law embody universal values that preserve lives and dignity.”

Two months later, the ICRC released its Challenges Report for 2024. It documented more than 120 active armed conflicts worldwide — involving more than 60 states and 120 non-state armed groups. It described systematic attacks on medical facilities, detainees subjected to torture, civilians treated as military targets. Non-compliance, the report concluded, remains “the single most important challenge for protecting people in armed conflict.” Not a new problem. Not an exceptional year. The persistent, ordinary baseline.

Both documents came from the same organization. Both were factually accurate. Read together, they describe the following situation: 196 states have ratified the most universally subscribed treaties in the history of international law, and the world currently has more than 120 active armed conflicts in which those treaties are being violated, and not one state official in a command position for those violations is in ICC custody. The 75th anniversary is therefore a genuine milestone and a precise indictment at the same time, depending on what question you’re asking.

The question worth asking is not why states violate these rules. States violate rules because rules without consequences are recommendations. That is not a mystery. The more interesting question is structural: how does a legal framework achieve near-universal ratification and near-systematic violation simultaneously, without this being understood as a design failure? The answer, when you follow the enforcement architecture back to its origins — through the Protecting Power system that no state has ever invoked, through the Fact-Finding Commission that has never conducted a full grave-breaches investigation, through the ICC’s jurisdictional map and the deliberate gaps in it — is that it might not be a failure at all.

What the Text Actually Says

The four Geneva Conventions of August 12, 1949 are a precise set of documents. Geneva Convention I covers the wounded and sick in land warfare. GC II covers the wounded, sick, and shipwrecked at sea. GC III covers prisoners of war — including Article 17, which limits what captors can demand (name, rank, date of birth, serial number) and Article 130, which enumerates grave breaches. GC IV covers civilians under occupation: prohibiting collective punishment, deportation, and destruction of civilian property.

Three Additional Protocols followed. AP I (1977) established the principles of proportionality and distinction in international armed conflict — the requirement to distinguish combatants from civilians and to avoid attacks whose harm to civilians is excessive relative to military advantage. AP II (1977) extended minimum protections to non-international armed conflicts. AP III (2005) added the Red Crystal as a recognized protective emblem.

Common Article 3 appears identically in all four Conventions. It sets minimum standards for civil wars and internal conflicts — protection of non-combatants, humane treatment, prohibition on torture and “outrages upon personal dignity.” In 1986, the International Court of Justice, in Nicaragua v. United States, confirmed that Common Article 3 reflects customary international law binding all states regardless of treaty ratification.

The grave breaches regime — Articles 50, 51, 130, and 147 across the four Conventions — is specific. Willful killing. Torture or inhuman treatment. Willful causing of great suffering. Unlawful deportation or transfer. Taking hostages. Compelling a prisoner of war to serve in the forces of the hostile power. These acts are not just prohibited; they are designated as triggering criminal accountability. States are required to search for persons alleged to have committed grave breaches and bring them before their own courts or hand them over for trial elsewhere.

196 states are party — more than have ratified the UN Charter. The rules are not ambiguous. The categories are defined. The gap between the text and the practice is not a product of vagueness.

It is worth pausing on how Common Article 3 got into the text at all. At the 1949 diplomatic conference, states resisted strenuously. Granting legal protections to non-state armed groups — rebels, insurgents, anti-colonial movements — meant implicitly recognizing them as parties with rights. The final text includes explicit language to address this fear: compliance “shall not affect the legal status of the Parties to the conflict.” States could follow the rules without conceding that the people they were fighting against had any legitimate standing. The sovereignty-preservation logic embedded in that single sentence previews the entire enforcement architecture that follows.

What a protecting power is supposed to do

GC I, Article 8 describes the Protecting Power system in some operational detail. A neutral state designated as Protecting Power would supervise the application of the Conventions, facilitate communication between belligerents, and — critically — visit prisoner of war camps and civilian internment centers to verify conditions. Article 8 of GC III (POWs) specifies that the ICRC may undertake the same functions. This is not an aspirational description. It is a defined operational role for a defined institutional actor. It has never been filled, by any state, in any conflict, since the Conventions entered into force. The gap is not between the text's aspirations and reality. It is between documented operational requirements and what states were ever willing to do.

The machinery that doesn’t run

The machinery that doesn’t run is best introduced by its failures.

Protecting Powers

No state has ever appointed a Protecting Power under the Geneva Conventions. Not for Korea. Not for Vietnam. Not for the Iran-Iraq War, the Falklands, the Gulf Wars, the Balkans conflicts of the 1990s, Afghanistan, Iraq, Syria, or Gaza. The system is specified in GC I through GC IV as a mandatory component of Convention implementation — each party to a conflict is required to appoint a neutral state to supervise application of the treaties from the start of hostilities. In practice, no party has ever done this.

During the Iran-Iraq War (1980-1988), both countries denied the ICRC access to prisoner of war camps. The Protecting Power mechanism, had it operated, would have provided precisely the kind of neutral verification that was absent. Instead, the fallback monitor was also excluded. Hundreds of thousands of soldiers spent years in captivity without any external verification of their treatment.

The International Humanitarian Fact-Finding Commission

The IHFFC was created under Article 90 of Additional Protocol I, and was formally constituted in 1991. Its mandate: to investigate alleged grave breaches and serious violations of IHL. The mechanism requires consent of the parties under investigation before the Commission can act.

In thirty-four years of existence, the Commission has been formally activated twice. In 2017, the OSCE asked the Commission to lead a forensic investigation in eastern Ukraine regarding an explosion that killed an OSCE paramedic and injured two monitors. The investigation concluded the incident likely resulted from a recently laid anti-tank mine — a finding with real forensic value, but not a grave-breaches investigation. In 2024, following the April 1 strike on World Central Kitchen aid workers in Gaza that killed seven people including a Polish national, the government of Poland signed a memorandum of understanding with the IHFFC regarding the incident. Poland was not a party to the conflict. The MOU created a framework for investigation without constituting a consent-based investigation by the parties responsible for the strike.

The Commission has never conducted a full grave-breaches investigation. Not once in three decades.

The ICRC’s Structural Trap

The ICRC is the only body with meaningful ongoing humanitarian access in most active conflicts. Its delegates visit detention facilities, monitor treatment of wounded combatants, trace missing persons. This access is real and it produces results. But its operating approach — “bilateral and confidential” — means that when delegates observe violations, they communicate findings privately to the responsible authorities. Nothing is published. No findings are disclosed.

The ICRC’s own explanation is honest about why: if parties believed the ICRC would make findings public, they would exclude the ICRC from access. A body known to publish what it sees cannot be granted the access that allows it to see anything. The organization chose monitoring over accountability, because monitoring was the only viable option — and because the alternative, an organization that published findings and was therefore denied access, would have neither information nor leverage.

There is no version of this trade-off that produces both. Either the ICRC stays silent and retains access to detainees, or the ICRC speaks and loses the rooms where detainees are held. The system is designed, by the parties to it, to produce exactly this choice. States grant access on the condition that findings remain private. The ICRC accepts, because the alternative is that the people in those rooms have no one. This is not a compromise that degraded over time. It is what was negotiated.

The ICRC can make public statements when private representations have been “exhausted and have proved fruitless.” It has done so, occasionally. But this option comes with a direct cost: it ends or severely limits access. The ICRC’s effectiveness as a monitoring body and its usefulness as an accountability mechanism are in direct structural tension. This is not a failure of institutional will. It is the logical consequence of what states were willing to permit.

Universal Jurisdiction

The Conventions require states to prosecute or extradite any person accused of grave breaches, regardless of the nationality of the accused or the location of the crime. The notable test case: Germany convicted Anwar Raslan, a Syrian intelligence colonel who served as head of the Investigations Division at Branch 251 of Syria’s General Intelligence Directorate, of crimes against humanity on January 13, 2022. The Koblenz Higher Regional Court found him responsible for 27 murders, 4,000 cases of torture, and three cases of sexual violence. He was sentenced to life imprisonment.

Raslan was mid-level. The architects of Syria’s detention policy — the Assad government, the military high command, the officials who designed and authorized the system of which Branch 251 was one node — were untouched. Not one P5 state has surrendered a senior military or government official for trial under universal jurisdiction. Universal jurisdiction operates at the margins of accountability. It reaches the reachable. It does not, by its nature and by the political reality of its operation, reach the people who design the systems.

The court that cannot reach

The ICC exercises jurisdiction over crimes committed on the territory of a member state, by a national of a member state, or — via a third pathway — through a UN Security Council referral that extends jurisdiction regardless of membership. The Rome Statute Articles 12 and 13 define these parameters.

The United States, Russia, and China are not members. Crimes committed by nationals of those three states, on the territory of non-member states, require a UNSC referral to fall within ICC jurisdiction. The veto power in the Security Council is held by five states: the US, Russia, China, the United Kingdom, and France. Three of the five states structurally outside ICC jurisdiction — and most capable of producing or shielding large-scale violations — collectively hold the power to prevent the mechanism that could create jurisdiction over them from ever being activated.

The logic is not coincidental. At the 1998 Rome Conference, Ambassador David Scheffer — the Clinton administration’s lead negotiator and the US head of delegation — stated clearly that the United States could not accept a treaty under which US forces could face ICC prosecution without US consent. This was not a late-stage hedge. It was the explicit American position throughout the negotiations. The US voted against the Rome Statute on July 17, 1998, one of seven states to do so. Clinton signed the treaty on December 31, 2000 — a move that preserved US ability to participate in the Court’s development without ratifying. The Bush administration “unsigned” in May 2002, formally notifying the UN that the US did not intend to ratify. It then negotiated Article 98 bilateral immunity agreements with more than 100 countries, ensuring that those states would not surrender US nationals to the Court.

The jurisdictional architecture shielding US forces was not a drafting oversight. It was the condition the US required, stated on the record by the US negotiator. Scheffer said so.

UNSC referrals have produced two ICC situations: Sudan/Darfur via Resolution 1593 in 2005 (Russia and China abstained), and Libya via Resolution 1970 in 2011 (same abstentions). Syria: vetoed. Yemen: no referral attempted. Multiple Security Council resolutions calling for a ceasefire in Gaza were vetoed by the United States throughout 2024.

The overrepresentation of African defendants in ICC proceedings is a documented pattern — the vast majority of those indicted and all convictions to date involve situations on the African continent. This is not evidence of selective targeting by the Office of the Prosecutor. It is the logical output of a system where membership, veto power, and Article 98 agreements determine who is reachable. The ICC investigates where it has jurisdiction. Jurisdiction is determined by who signed, who didn’t, and who holds the veto. The result is a court that has convicted thirteen people, all in African situations, in twenty-five years of operation, during which European states conducted occupation warfare in Iraq and Afghanistan, produced documented torture programs, and faced no ICC proceedings that resulted in charges.

The warrants that cannot be served

In March 2023, the ICC issued an arrest warrant against Vladimir Putin for the war crime of unlawfully deporting Ukrainian children. In November 2024, it issued warrants against Israeli Prime Minister Benjamin Netanyahu and former Defense Minister Yoav Gallant for war crimes and crimes against humanity in Gaza. These are genuine departures from the historical pattern. The Court acted against a sitting leader of a P5 state and against the leader of a close US ally. Neither Israel nor Russia would surrender subjects for trial. The warrants generated immediate political counter-pressure on the Court itself: in February 2025, the Trump administration signed an executive order imposing sanctions on ICC officials, targeting judges and prosecutors working on the Gaza case. Subsequent rounds of sanctions in 2025 added more judges and deputy prosecutors to the list. The warrants are real. They carry legal weight and produce political consequences. But the enforcement mechanism for executing them is the voluntary compliance of states — and the states where the subjects reside have no intention of complying.

The United States as worked example

The United States has not ratified Additional Protocol I. The Reagan administration formally declined in 1987. Legal Adviser Abraham Sofaer identified two provisions as “fundamentally and irreconcilably flawed.” Article 1(4) would extend combatant protections to fighters in wars of national liberation against colonial domination or foreign occupation — which the administration read as extending legal status to the PLO and similar organizations. Article 44(3) modifies the distinction requirement in ways the Joint Chiefs assessed as “militarily unacceptable,” relaxing the requirement for combatants to carry arms openly in certain circumstances. These were not abstract concerns. They were specific objections to specific provisions that would have constrained specific operational choices.

The AP I non-ratification is the predicate to everything that follows. Proportionality and distinction are now recognized as customary international law binding even on non-parties — the ICRC Customary IHL Study (Henckaerts and Doswald-Beck, 2005) identifies 161 rules of customary IHL, drawing them across both treaty and non-treaty states. But the US formally contested this study’s conclusions, arguing that the ICRC mischaracterized treaty-specific rules from AP I as customary law binding on states that never ratified the Protocol. The US and the ICRC have ongoing formal disagreement about how much law is actually binding on states that signed the core Conventions but not the Protocols. The customary-law back-door to enforcement is itself contested by the state most capable of blocking formal enforcement.

Abu Ghraib documented what domestic accountability looks like in practice. Between 2003 and 2004, US military personnel at the Abu Ghraib prison complex in Iraq systematically subjected detainees to torture, sexual humiliation, and degrading treatment. Manadel al-Jamadi died in US custody at the facility — beaten during interrogation, his body documented in the now-familiar photographs. Eleven US soldiers were convicted of related offenses. Lynndie England received a three-year sentence and was released on parole after approximately eighteen months. Charles Graner, identified as a ringleader, received ten years and served six and a half.

None of this was disputed. The photographs made denial impossible. The question was always how far up the chain accountability would travel.

It stopped at the enlisted level. Donald Rumsfeld, who authorized expanded interrogation techniques that created the conditions in which Abu Ghraib occurred, faced no criminal proceedings. Jay Bybee and John Yoo, the Justice Department lawyers who authored the Office of Legal Counsel memoranda redefining torture to exclude techniques that fell short of organ failure or death — legal arguments later repudiated by the OLC itself — were never charged. CIA contractors involved in the program were not prosecuted. Human Rights Watch’s 2004 report “Road to Abu Ghraib” traced the policy chain from Rumsfeld’s directives to the prison floor. HRW documented in 2023 that two decades later, Iraqi survivors received no compensation and no prosecutorial accountability had reached above the enlisted rank.

The state that committed the violations controlled who was charged, what evidence was gathered, and how far up the chain any investigation traveled. What kept it at the enlisted level was the framework’s fundamental design: domestic prosecution, no external mechanism, complete state control over the scope of accountability.

Yemen offers the same structure with different actors. Saudi coalition forces struck hospitals whose GPS coordinates had been shared with coalition commanders — an act documented by both Human Rights Watch and Médecins Sans Frontières in 2015 and 2016. The United States provided military support to the coalition, including intelligence and logistics. No ICC referral was possible. No accountability mechanism was available or attempted. The architecture of non-accountability does not require US forces to be the direct perpetrators. It operates wherever the protections are structurally absent.

What the system actually achieves

Absent prosecution, the system still produces something.

States modify behavior when violations are documented, attributed, and publicized. Not from fear of prosecution — there is no prosecution to fear for the powerful — but because reputational costs are real. The ICRC’s bilateral and confidential approach produces results in individual cases: POW access negotiated, detention conditions improved, specific prisoners located. MSF’s public documentation of hospital strikes in Yemen generated political pressure that domestic accountability mechanisms could not. The mechanism fails when states decide the benefit of the violation exceeds the reputational cost. It fails regularly. But it is not nothing.

The substantive law has also spread beyond its treaty form. Nicaragua confirmed Common Article 3 as customary law. AP I’s core principles — proportionality, distinction, precaution — are now recognized as customary despite the US non-ratification. The ICRC Customary IHL Study formalized this. States that never ratified AP I are still legally bound by its core prohibitions, in the view of most international law scholars and the ICRC. This is a genuine expansion of the normative framework even as enforcement remains structurally weak. The US contesting the customary study is itself evidence of how seriously states take it.

Then there are the tribunals. Nuremberg. The International Criminal Tribunal for the former Yugoslavia. The International Criminal Tribunal for Rwanda. The Special Court for Sierra Leone. Germany’s conviction of Anwar Raslan. All of these drew on the Geneva Conventions’ definitions of war crimes and crimes against humanity. None operated through the Conventions’ own enforcement mechanisms. All required separate instruments — post-conflict political will, ad hoc tribunal creation, domestic prosecutorial initiative. The Conventions function as a repository of norms that can be activated when the political conditions exist for activation. Those conditions typically materialize when a powerful actor has already lost, or when a sufficiently determined third-country court acts against a defendant who can be reached.

The distinction matters: the ICTY prosecuted Bosnian Serb and Croatian officials because the political constellation in the early 1990s made ad hoc tribunal creation possible — the US, Russia, and European powers all had reasons to support it. The ICTR prosecuted Rwandan genocide perpetrators because the new Rwandan government consented and the international community was visibly ashamed of its inaction. Germany prosecuted Raslan because he was physically on German territory, and the German judiciary was willing. These are real accountability events. They are also illustrations of what it takes to produce accountability: a specific political alignment that will not be repeated on demand.

This is genuinely something. Raslan spent the rest of his life in German prison for what he did in Damascus. The ICTY convicted Radovan Karadzic. Nuremberg executed Nazi leadership. These outcomes happened. The norms encoded in the Conventions made them possible. It is just not enforcement in the sense that enforcement is usually understood — as a reliable, predictable consequence of violation. It is contingent accountability: accountability that materializes in specific circumstances, for specific actors, when specific conditions align. For the architects and the powerful, those conditions very rarely align.

Closing

Return to August 12, 2024. The celebration was accurate in a narrow sense. 196 states have ratified the same four documents. That is a genuine achievement of international legal history — more states formally bound by a set of norms than any other treaty framework in existence.

But states ratified because the enforcement architecture preserved their sovereignty. They got the reputational benefit — the moral standing of formal commitment, the right to cite the Conventions when condemning adversaries, the sense of participation in a civilized international order — without surrendering practical control over whether their own conduct would ever be adjudicated. Every mechanism designed to create external accountability was either never used or structured to require the consent of the violator. Protecting Powers have never been invoked — invoking one means accepting a neutral monitor with actual access. The IHFFC has never conducted a full grave-breaches investigation — triggering it means consenting to scrutiny. The ICRC operates in silence because the alternative is exclusion. The ICC cannot reach the most powerful states because the most powerful states designed it not to reach them, and Ambassador Scheffer said so at the conference, on the record, as the explicit US negotiating position.

This is what the laws of armed conflict look like when they reflect the maximum that states were actually willing to agree to, rather than the minimum necessary to protect the people war most reliably destroys. The 75th anniversary and the Challenges Report are both accurate. They are describing the same system from different vantage points. The system is functioning. It is functioning in accordance with what powerful states were willing to build.

Gen AI Disclaimer

Some contents of this page were generated and/or edited with the help of a Generative AI.

Media

Facsimile of the original first Geneva Convention from 1864, on loan to the International Red Cross and Red Crescent Museum in Geneva, Switzerland. – Wikipedia

Key Sources and References

International Committee of the Red Cross. “Geneva Conventions 75th Anniversary: Foundational Treaties Save Lives and Dignity, but Massive Humanitarian Suffering Shows the World Must Recommit.” Press release, August 12, 2024. icrc.org.

International Committee of the Red Cross. “International Humanitarian Law and the Challenges of Contemporary Armed Conflicts: Building a Culture of Compliance for IHL to Protect Humanity in Today’s and Future Conflicts.” Challenges Report, October 2024. icrc.org.

International Committee of the Red Cross. “IHL Treaties — Treaties and States Parties.” ihl-databases.icrc.org. (196 state parties to the 1949 Conventions confirmed.)

Henckaerts, Jean-Marie, and Louise Doswald-Beck. Customary International Humanitarian Law. Volume I: Rules. Cambridge University Press / ICRC, 2005.

International Court of Justice. Case Concerning Military and Paramilitary Activities in and Against Nicaragua (Nicaragua v. United States of America). Judgment of June 27, 1986. ICJ Reports 1986. icj-cij.org.

International Humanitarian Fact-Finding Commission. 2017 Ukraine Investigation Report (OSCE-initiated forensic investigation). ihffc.org.

International Humanitarian Fact-Finding Commission. Memorandum of Understanding with Poland regarding the April 1, 2024 WCK incident in Gaza, signed June 20, 2024. ihffc.org.

Koblenz Higher Regional Court. Judgment in the case of Anwar Raslan. January 13, 2022. Summarized by European Center for Constitutional and Human Rights (ECCHR). ecchr.eu.

Human Rights Watch. “Road to Abu Ghraib.” June 2004. hrw.org.

Human Rights Watch. “Iraq: Torture Survivors Await US Redress, Accountability.” September 25, 2023. hrw.org.

Human Rights Watch. “Saudi Arabia/Yemen: Unlawful Coalition Strikes on Hospitals.” December 2015. hrw.org.

Médecins Sans Frontières. “Documents: MSF’s Interaction with Parties to the Conflict in Yemen.” 2016. msf.org.

Scheffer, David L. Statement before the Senate Foreign Relations Committee on the International Criminal Court. July 23, 1998. 105th Congress, 2nd Session. govinfo.gov (CHRG-105shrg50976).

International Criminal Court. “About the Court.” icc-cpi.int/about/the-court. (Thirteen convictions and four acquittals as of October 2025.)

International Criminal Court. Warrant of Arrest for Vladimir Putin. March 17, 2023. icc-cpi.int.

International Criminal Court. Warrants of Arrest for Benjamin Netanyahu and Yoav Gallant. November 21, 2024. icc-cpi.int.

White House. Executive Order 14203: “Imposing Sanctions on the International Criminal Court.” February 6, 2025. whitehouse.gov.

Reagan, Ronald. Letter of Transmittal declining Additional Protocol I. January 29, 1987. Published with Legal Adviser Abraham Sofaer’s analysis. US Department of State.

Bouchet-Saulnier, Françoise. “Protecting Powers.” The Practical Guide to Humanitarian Law. Médecins Sans Frontières. guide-humanitarian-law.org.

ICRC. Casebook: “ICJ, Nicaragua v. United States.” casebook.icrc.org. (Common Article 3 as customary IHL.)