International Arbitration and the New York Convention: Enforcing Awards Across Borders

Document type: Article Practice area: Litigation — Arbitration Jurisdiction: United States and international Last reviewed: 5 September 2026

Here is a fact that surprises people who do not practice in this area: it is materially easier to enforce a Singaporean arbitral award in a United States court than a Singaporean court judgment.

The judgment depends on state recognition statutes, comity, and a set of doctrines that vary by state and by the relationship between the two legal systems. The award depends on a treaty with more than 170 contracting states, a short list of defenses, and a strong presumption of enforcement. The award usually wins.

That asymmetry is not an accident. The Convention on the Recognition and Enforcement of Foreign Arbitral Awards — the New York Convention, concluded in 1958 — was designed to make arbitral awards portable in a way judgments are not, and it worked. It is routinely described as the most successful private international law instrument of the twentieth century, and the description is not hyperbole. Cross-border commerce is arbitrated because of it.

For lawyers, the Convention is short — sixteen articles, and only two of them do most of the work — and the practical consequences are large. Understanding it is the difference between an award that turns into money and an award that turns into a decade of litigation in four jurisdictions.

The architecture

Chapter 2 of the FAA, 9 U.S.C. §§ 201–208, implements the Convention in the United States. Chapter 3, §§ 301–307, implements the Inter-American Convention on International Commercial Arbitration — the Panama Convention — which governs among most Latin American states.

Which awards fall under the Convention? Section 202 supplies a definition broader than "foreign award" suggests. An arbitration agreement or award arising out of a commercial legal relationship falls under the Convention unless it is entirely between citizens of the United States. And even an agreement or award entirely between U.S. citizens falls under it if the relationship "involves property located abroad, envisages performance or enforcement abroad, or has some other reasonable relation with one or more foreign states."

The practical effect: a great many arbitrations that feel domestic are Convention arbitrations. A dispute between two Delaware corporations over a supply contract performed in Vietnam is a Convention case. So is a dispute between a U.S. company and the U.S. subsidiary of a foreign parent, if the subsidiary is a foreign citizen under § 202's rule that a corporation is a citizen of the United States only if it is incorporated here or has its principal place of business here.

Why this matters immediately. Convention status changes three things:

Jurisdiction. Section 203 confers original federal district court jurisdiction over any action falling under the Convention, "regardless of the amount in controversy." This is a genuine federal question grant — unlike Chapter 1, where the FAA supplies no jurisdiction and Badgerow v. Walters, 596 U.S. 1 (2022) forecloses looking through a confirmation petition to the underlying dispute. Chapter 2 has no such problem.

Removal. Section 205 permits removal of a state court action that "relates to" a Convention arbitration agreement or award, "at any time before the trial thereof." This is an unusually generous removal provision: no thirty-day limit, no unanimity requirement in most readings, and a "relates to" standard courts read broadly.

Time. Section 207 gives three years to seek confirmation, not the one year of § 9.

Article V: the seven defenses

Section 207 directs the court to confirm a Convention award "unless it finds one of the grounds for refusal or deferral of recognition or enforcement of the award specified in the said Convention." Those grounds are in Article V, and there are seven of them — five that the resisting party must prove, and two the court may raise itself.

Article V(1) — the party-invoked defenses.

(a) Incapacity or invalidity. A party to the agreement was under some incapacity, or the agreement is not valid under the law to which the parties subjected it or, failing any indication, the law of the country where the award was made.

(b) Lack of notice or inability to present the case. The party was not given proper notice of the appointment of the arbitrator or of the proceedings, or was otherwise unable to present its case. This is the due process defense, and it is the most commonly asserted. Courts read it narrowly: the standard is a fundamental inability to be heard, not dissatisfaction with procedural rulings.

(c) Excess of scope. The award deals with a difference not contemplated by or not falling within the terms of the submission, or contains decisions on matters beyond the scope of the submission. Note the severability proviso: if the decisions on submitted matters can be separated, that part may be enforced.

(d) Irregular composition or procedure. The composition of the tribunal or the arbitral procedure was not in accordance with the parties' agreement, or, failing agreement, with the law of the country where the arbitration took place.

(e) Award not binding, or set aside at the seat. The award has not yet become binding, or has been set aside or suspended by a competent authority of the country in which, or under the law of which, that award was made.

Article V(2) — the court-raised defenses.

(a) Non-arbitrability. The subject matter is not capable of settlement by arbitration under the law of the enforcing country.

(b) Public policy. Recognition or enforcement would be contrary to the public policy of that country.

How narrow are these in practice? Very. United States courts have consistently held that the public policy defense applies only where enforcement would violate "the forum state's most basic notions of morality and justice." Allegations that the tribunal misapplied the law, misread the contract, or reached an unjust result do not qualify. The defense succeeds in cases involving genuine corruption in the procurement of the award, or awards that would require conduct illegal in the enforcing state — not in cases where a party is unhappy.

The Supreme Court's international arbitration cases set the tone. Scherk v. Alberto-Culver Co., 417 U.S. 506 (1974) enforced an international arbitration clause over a securities claim, emphasizing that international commerce requires predictability about forum. Mitsubishi Motors Corp. v. Soler Chrysler-Plymouth, Inc., 473 U.S. 614 (1985) held antitrust claims arbitrable in the international context, reasoning that "concerns of international comity, respect for the capacities of foreign and transnational tribunals, and sensitivity to the need of the international commercial system for predictability in the resolution of disputes" require enforcing the agreement. Both decisions are cited constantly, and both point the same direction.

The seat: the most important word in the clause

Practitioners new to international arbitration frequently confuse the seat with the hearing venue, and the confusion produces real damage.

The seat — sometimes called the legal place or juridical seat — is the arbitration's legal home. It determines:

Which country's arbitration law governs the procedure (the lex arbitri).
Which courts have supervisory jurisdiction — to appoint arbitrators, to grant interim relief, and, critically, to set the award aside.
Whether the award is "foreign" or "domestic" for Convention purposes in a given enforcing state.
Which mandatory procedural rules apply.

The hearing venue is where people physically meet. It has no legal significance and can be anywhere convenient.

A clause that says "arbitration shall take place in Geneva" is ambiguous — is Geneva the seat or the venue? Most tribunals will read it as the seat, but the ambiguity itself becomes an issue. A clause that says "the seat of the arbitration shall be Geneva, Switzerland; hearings may be held at any convenient location" is clear.

Why the seat matters most. Under Article V(1)(e), an award set aside at the seat may be refused enforcement everywhere. The seat's courts hold a power no other court holds: they can kill the award at its source. Choosing a seat is therefore choosing which national judiciary supervises the arbitration, and the differences among candidate seats are real.

What makes a good seat:

A pro-arbitration judiciary with a demonstrated record of narrow review and prompt decisions.
A modern arbitration law, typically based on the UNCITRAL Model Law.
Convention membership — essential.
Neutrality relative to the parties.
Practical infrastructure — counsel, arbitrators, interpreters, hearing facilities, though this matters less in an era of remote hearings.
No unusual mandatory rules that override party autonomy.

The commonly chosen seats — London, Paris, Geneva, Singapore, Hong Kong, New York, Stockholm — earned their positions by combining most of these. A seat chosen for a party's convenience rather than for these qualities is a decision that can look expensive years later.

Institutions and rules

The choice of administering institution is separate from the seat and separate from the governing law.

The ICC International Court of Arbitration is the most widely used institution for large cross-border commercial disputes. Distinguishing features: the Terms of Reference, a document defining the issues that the tribunal and parties sign early; and scrutiny of the award by the ICC Court before it issues, which catches errors and improves enforceability. Both add time. Both add value in large cases.

The LCIA (London Court of International Arbitration) is leaner and less interventionist, with a strong reputation for efficient administration.

The ICDR, the international division of the AAA, administers a large volume of U.S.-connected international cases with rules familiar to American practitioners.

SIAC (Singapore) and HKIAC (Hong Kong) have grown rapidly and are the default choices for many Asia-Pacific transactions, both offering expedited procedures and emergency arbitrator provisions.

UNCITRAL Rules govern ad hoc arbitrations with no administering institution. They work, but they place more burden on the parties and the tribunal, and they require careful attention to the appointing authority. For most commercial parties, institutional administration is worth its cost.

What the choice actually affects: the appointment mechanism, the availability of emergency and expedited procedures, the consolidation and joinder provisions, the cost structure (some institutions charge ad valorem, some hourly), the confidentiality default, and the degree of institutional oversight of the award.

Awards annulled at the seat

Article V(1)(e) permits an enforcing court to refuse enforcement of an award "set aside or suspended by a competent authority of the country in which, or under the law of which, that award was made."

Note the word: may. The English text of Article V(1) says recognition and enforcement "may be refused" on the enumerated grounds. It does not say must.

This has produced one of the genuinely contested questions in the field. Some jurisdictions — France most prominently — have enforced awards annulled at the seat, reasoning that an international award is not anchored to any national legal order and that a seat court's annulment is a local act with local effect. Other jurisdictions treat annulment as effectively dispositive.

United States courts have enforced annulled awards in a small number of cases, generally requiring a showing that the annulment offended basic standards of justice — that it was procured by corruption, that it applied a legal standard repugnant to U.S. public policy, or that it was rendered by a judiciary lacking independence. The threshold is high, and most annulled awards are not enforced here.

What this means practically. The seat's supervisory jurisdiction is powerful but not absolute. A party facing annulment proceedings at a seat whose courts are compromised has an argument elsewhere. A party facing annulment at a respected seat, on ordinary grounds, has very little.

And it means the seat choice matters even more. Choosing a seat whose courts are unpredictable is choosing a risk that surfaces only at the end, when it is most expensive.

Discovery: the two systems, and what ZF Automotive changed

International arbitration does not do American discovery. This is one of the most consistent sources of client surprise, and it cuts both ways.

The baseline. Most international arbitrations follow something like the IBA Rules on the Taking of Evidence in International Arbitration, which are a compromise between common law and civil law traditions. Under them:

Each side submits the documents it relies on with its written submissions.
A party may make a narrow and specific document request — the Redfern Schedule format — identifying a document or a narrow category, explaining its relevance and materiality, and stating why it is presumed to be in the other side's possession.
The tribunal rules request by request.
Depositions are rare to nonexistent.
Interrogatories do not exist.
Witness evidence comes in written statements, with cross-examination at the hearing.
Experts submit reports and are frequently required to confer and produce a joint statement of agreement and disagreement before the hearing.

What this means for an American party. The instinct to serve broad requests will be rejected, and repeated broad requests damage credibility with the tribunal. The discipline is to identify the ten or fifteen documents that actually matter and request those precisely.

What it means for a non-U.S. party. There is more disclosure than a civil law practitioner expects, and adverse inferences from a failure to produce are real. Article 9(5) of the IBA Rules permits the tribunal to infer that a withheld document would be adverse to the withholding party, and tribunals use it.

Section 1782 and its limits

28 U.S.C. § 1782 permits a U.S. district court to order a person residing or found in the district to give testimony or produce documents "for use in a proceeding in a foreign or international tribunal." For years, parties used it to obtain U.S.-style discovery in aid of foreign arbitrations, and the circuits split on whether a private arbitral tribunal counted.

ZF Automotive US, Inc. v. Luxshare, Ltd., 596 U.S. 619 (2022) resolved it. The Court held that only a governmental or intergovernmental adjudicative body qualifies as a "foreign or international tribunal" under § 1782. A private commercial arbitral panel — a DIS panel in the ZF Automotive case itself — does not. The companion case involved an ad hoc panel constituted under a bilateral investment treaty, and the Court held that it too fell outside § 1782 because the treaty parties had not clothed it with governmental authority.

The practical consequences:

Section 1782 is unavailable for ordinary international commercial arbitration. Parties who built strategies around it have had to rebuild them.
It remains available for foreign court proceedings, and for genuinely governmental or intergovernmental bodies.
Investor-state arbitration is fact-dependent after ZF Automotive. Some tribunals may qualify depending on how the treaty constitutes them; most private-panel arrangements will not.
The remaining tool is 9 U.S.C. § 7, which permits arbitrators to summon witnesses and documents — but it applies to domestic arbitrations under Chapter 1, and its reach to non-parties and its geographic limits are themselves contested.

What sophisticated parties do instead. They negotiate document production expressly in Procedural Order No. 1, they use the Redfern Schedule aggressively but precisely, and they rely on adverse inferences. In transactions where discovery will matter — a fraud claim, an earn-out dispute turning on the buyer's internal records — they negotiate document production obligations into the arbitration clause itself.

Interim relief and emergency arbitrators

A party that needs to freeze an asset or preserve evidence before a tribunal is constituted has two routes, and it usually needs both.

The emergency arbitrator. Every major institution now provides one: the ICC, LCIA, ICDR, SIAC, and HKIAC rules all permit a party to apply for emergency relief before the tribunal is formed, typically with an arbitrator appointed within one to two business days and a decision within about two weeks. The relief takes the form of an order or award and binds the parties contractually.

The problem with emergency relief: it does not bind third parties, and its enforceability as an "award" under the Convention is contested. A bank holding the asset is not a party to the arbitration agreement.

Court-ordered interim relief. The Convention does not prohibit a party from seeking interim measures from a court, and most jurisdictions permit it. In the United States, courts have generally held that seeking interim relief in aid of arbitration is not a waiver of the arbitration agreement, though the circuits have expressed the point differently.

The practical sequence for a party facing dissipation of assets:

File the arbitration — many institutional rules require a pending or imminent case for emergency relief.
Apply for an emergency arbitrator for relief binding the counterparty.
Simultaneously seek court relief where the assets are, for relief binding third parties.
Draft for it in advance. A clause that expressly preserves the right to seek interim relief from any court of competent jurisdiction removes an argument the counterparty will otherwise make.

Sovereign parties and investor-state arbitration

Where the counterparty is a state or a state-owned entity, the analysis changes.

Immunity from suit. Under the Foreign Sovereign Immunities Act, a foreign state is presumptively immune. The arbitration exception — which withdraws immunity where an action is brought to enforce an agreement to arbitrate or to confirm an award governed by a treaty such as the New York Convention — is the usual route, and it is generally effective for confirmation proceedings.

Immunity from execution is separate and harder. Confirming an award against a state produces a judgment; collecting it requires finding property in the United States used for a commercial activity, and central bank assets, diplomatic property, and military property are protected. Parties routinely obtain judgments against states that they cannot collect, and enforcement campaigns run for years across multiple jurisdictions.

Investor-state arbitration — claims by investors against states under bilateral or multilateral investment treaties — is a distinct field with its own institutions (principally ICSID), its own procedural framework, and its own annulment regime. ICSID awards under the ICSID Convention are enforced through a separate mechanism that does not run through the New York Convention at all: contracting states must recognize ICSID awards as if they were final judgments of their own courts. Non-ICSID investment awards, including those under the UNCITRAL Rules, are enforced under the Convention like commercial awards.

Practical note: contracting with a state-owned entity is not the same as contracting with the state, and whether the entity is an "agency or instrumentality" for FSIA purposes is a fact question that should be evaluated before the contract is signed, not after the award.

A worked enforcement campaign

Brannoch Marine Systems, a Norwegian manufacturer, obtained an ICC award of €14.2 million against Palgrave Offshore Holdings, a Cyprus company with operating subsidiaries in Texas, the Netherlands, and Singapore. The seat was London. The award issued in March.

Brannoch's counsel, Ingrid Achebe-Solheim, ran a campaign rather than a case.

Step one: map the assets before filing anything. Three months of investigation — corporate registries, vessel registries, shipping databases, credit reports, and a forensic accountant — produced an asset map: two supply vessels registered in Panama but trading in the Gulf of Mexico, receivables from three U.S. operators, a Rotterdam warehouse lease with a security deposit, and a Singapore bank relationship.

Step two: pick the order of jurisdictions. She filed first in the Southern District of Texas, because the vessels called there and because § 203 gave automatic federal jurisdiction with no diversity or amount question. She did not file first in England, the seat, because filing at the seat would have invited a set-aside application she preferred Palgrave to have to initiate.

Step three: the U.S. petition. Filed under §§ 207 and 203, attaching the award, the agreement, and certified translations. Palgrave opposed on Article V(1)(b), arguing it had been unable to present its case because the tribunal denied a request to add a technical expert late in the proceeding.

The court rejected it in eleven weeks. The tribunal had given Palgrave notice, a hearing, the opportunity to submit expert evidence within the schedule, and a reasoned explanation of the denial. Article V(1)(b) addresses fundamental due process, not disagreement with a scheduling ruling.

Step four: the vessel arrest. Concurrent with confirmation, Brannoch arrested one supply vessel when it called at Galveston. The arrest produced a negotiation within eleven days, because a vessel under arrest is a vessel not earning.

Step five: parallel proceedings. Recognition applications were filed in the Netherlands and Singapore, both Convention states with efficient enforcement procedures. The Dutch proceeding attached the warehouse deposit; the Singapore proceeding produced a garnishment order against the bank relationship.

Step six: the seat. Palgrave did apply to set aside in London, on a challenge to the tribunal's jurisdiction over one respondent. The English court dismissed it seven months later. Had the application succeeded, Article V(1)(e) would have given every enforcing court a basis to refuse — which is why Ingrid tracked it as the campaign's principal risk throughout.

Outcome: €11.6 million recovered over nineteen months, at an enforcement cost of roughly €1.4 million across four jurisdictions.

Three lessons. First, the asset map comes before the filing. Three months of investigation directed the entire campaign. Second, enforcement is parallel, not sequential — filing in one jurisdiction and waiting produces years of delay. Third, the seat is the single point of failure, and it should be monitored from the first day of the enforcement campaign, not discovered when a set-aside order arrives.

Cost and duration

Duration. A mid-size ICC arbitration — €5 million to €50 million in dispute, three arbitrators — typically runs eighteen to thirty months from request to award. Expedited procedures, available at most institutions below a monetary threshold, compress this to six to twelve months with a sole arbitrator.

Costs come in three buckets:

Institutional fees, which at the ICC are ad valorem on the amount in dispute and can be substantial in large cases.
Arbitrator fees, ad valorem at the ICC and hourly at the LCIA and most other institutions. Three arbitrators cost roughly three times one, and the decision to use a three-member tribunal should be made on the value and complexity of the dispute, not by default.
Counsel and expert fees, which dominate — typically sixty to eighty percent of the total.

Cost allocation. International tribunals allocate costs far more readily than American courts. The prevailing party frequently recovers a substantial portion of its legal fees, and tribunals consider conduct — obstructive document practice, abandoned claims, unreasonable settlement positions — in the allocation. This changes settlement dynamics meaningfully compared to American litigation, and clients should be told about it at the outset.

Third-party funding is available and increasingly common. Disclosure obligations vary: some institutions and some seats now require disclosure of funding arrangements, principally to allow conflict checks against the funder. Check the applicable rules before entering a funding agreement.

Who decides jurisdiction in an international arbitration

The question of who decides whether the tribunal has authority — the tribunal or a court — arises in every contested international case, and the international answer differs in emphasis from the domestic one.

Competence-competence. Nearly every modern arbitration law and institutional rule empowers the tribunal to rule on its own jurisdiction. The UNCITRAL Model Law provides for it expressly; so do the ICC, LCIA, ICDR, SIAC, and HKIAC rules. This is the principle of Kompetenz-Kompetenz, and it is close to universal.

Separability. The arbitration clause is treated as an agreement separate from the contract containing it, so an attack on the contract as a whole does not deprive the tribunal of authority. The United States reached the same result in Prima Paint Corp. v. Flood & Conklin Manufacturing Co., 388 U.S. 395 (1967) and reaffirmed it in Buckeye Check Cashing, Inc. v. Cardegna, 546 U.S. 440 (2006). The doctrine is essentially the same internationally.

But competence-competence is not exclusive. In most systems, including the U.S. system, a tribunal's jurisdictional ruling is subject to some judicial review — at the seat on a set-aside application, and at enforcement under Article V(1)(a) and V(1)(c). The tribunal decides first; the courts decide finally.

Treaty preconditions. BG Group plc v. Republic of Argentina, 572 U.S. 25 (2014) addressed a treaty provision requiring an investor to litigate in local courts for eighteen months before arbitrating. BG Group did not, and the tribunal excused the requirement. The Supreme Court treated the local-litigation requirement as a procedural precondition for the arbitrators to interpret, not a jurisdictional condition for courts, and upheld the award. The framework — conditions precedent to arbitration are presumptively for arbitrators; questions of whether the parties agreed to arbitrate at all are presumptively for courts — travels well beyond investment treaties.

Non-signatories. GE Energy Power Conversion France SAS v. Outokumpu Stainless USA, LLC, 590 U.S. 432 (2020) held that the New York Convention does not prohibit enforcement of an arbitration agreement by or against a non-signatory under domestic equitable estoppel doctrines. The Convention sets a floor, not a ceiling: it requires enforcement of written agreements between signatories and does not forbid a contracting state from going further. This matters in supply chains and project structures where the party you need to arbitrate with did not sign the contract you are suing on.

Practical differences from U.S. litigation

Clients experienced only in American courts consistently misjudge four things.

Written submissions carry the weight. The memorial system — long, comprehensive written submissions with all supporting evidence attached — replaces the American sequence of pleading, discovery, and motion practice. The first memorial is often the most important document in the case, and it is prepared over months, not weeks.

There are no dispositive motions in the American sense. Some rules now provide for early determination of manifestly unmeritorious claims, and tribunals use it sparingly. A party expecting summary judgment will not get it.

Witness evidence is written first. Direct testimony is submitted as a signed witness statement. The hearing is cross-examination. This makes hearings shorter and makes witness preparation different — the statement is the testimony, and it is drafted with counsel, which is permitted and expected in most international practice but bounded by professional rules and by the IBA Guidelines on Party Representation.

The tribunal is active. International arbitrators ask questions, direct the parties to address specific issues, and often circulate a list of questions before the hearing. A team that prepares only its own case and not the tribunal's questions is unprepared.

Drafting decisions that determine enforceability

Every enforcement problem in this article traces back to the clause. Nine decisions matter.

One — the seat, named as the seat. "The seat of the arbitration shall be [city, country]." Then separately: "Hearings may be held at any location the tribunal determines, without affecting the seat."

Two — the institution and rules, named precisely. "Administered by the International Chamber of Commerce under its Rules of Arbitration in effect at the time of the request." Vague references — "arbitration under international rules" — produce a fight about which rules before the case starts.

Three — the number of arbitrators, and how they are chosen. One arbitrator for smaller disputes; three where the amounts justify it. Specify the appointment mechanism and a deadline.

Four — the language. Name it. A clause silent on language in a contract between parties with different languages produces a preliminary dispute and a translation budget.

Five — the governing law of the contract, and separately of the arbitration agreement. These can differ, and where they do, the law of the arbitration agreement governs its validity and scope. Most clauses address only the first, and the omission has produced litigation in several jurisdictions. Add: "The law governing this arbitration agreement shall be the law of [X]."

Six — scope, drawn broadly. "Any dispute, controversy, or claim arising out of, relating to, or in connection with this Agreement, including its formation, validity, breach, termination, or invalidity."

Seven — interim relief preserved. "Nothing in this clause prevents a Party from seeking interim or conservatory measures from any court of competent jurisdiction, and such an application shall not be deemed incompatible with or a waiver of this agreement to arbitrate."

Eight — document production, if it will matter. Where a foreseeable dispute will turn on the other side's internal records, say so: "The IBA Rules on the Taking of Evidence in International Arbitration shall guide the tribunal's determination of document production requests."

Nine — confidentiality, stated. Institutional rules differ on whether arbitration is confidential by default. Do not assume; provide for it, with carve-outs for enforcement, legal compliance, and disclosure to insurers, auditors, and lenders.

And one thing to leave out: expanded judicial review of the award. It does not work under the FAA, and it does not travel well internationally either. If a second look matters, use an institutional appellate procedure where one exists, or accept the finality that is the point of the system.

The Panama Convention and the overlap problem

Chapter 3 of the FAA, 9 U.S.C. §§ 301–307, implements the Inter-American Convention on International Commercial Arbitration, concluded at Panama City in 1975 and generally called the Panama Convention. Its substantive provisions closely track the New York Convention — § 302 incorporates several New York Convention sections by reference — with one notable difference: absent party agreement on procedure, the rules of the Inter-American Commercial Arbitration Commission apply by default.

When does Panama displace New York? Section 305 supplies the rule: where a majority of the parties to the arbitration agreement are citizens of states that have ratified the Panama Convention and are members of the Organization of American States, Panama applies. Otherwise New York applies.

Why anyone cares. In practice the two Conventions produce similar outcomes, and the overlap question rarely changes a result. It does change which defenses are cited and which default rules fill gaps, and a party drafting a clause with Latin American counterparties should know which regime it is in. Several states — Brazil, Mexico, Argentina, Chile, Colombia, and others — are party to both.

What can go wrong, and how it usually does

The clause names a city and nothing else. "Arbitration in Zurich" leaves the institution, the rules, the number of arbitrators, the language, and the law of the arbitration agreement all open. Every one of them becomes a preliminary dispute.

The award is unenforceable where the assets are. A party wins in a Convention state and discovers the debtor's assets sit in a non-Convention state, or in a state whose courts apply Article V(2)(b) expansively. Check enforcement jurisdictions before drafting, not after winning.

The set-aside application is lost by inattention. The seat's courts operate on the seat's timetable, with the seat's procedural requirements, and a party defending an award abroad must engage local counsel at the seat immediately when an application is filed.

Translations are done badly or late. Enforcement courts require certified translations of the award and the agreement. Budget for them, and use translators experienced in legal documents; a mistranslation of the operative award language creates an argument that did not exist.

The corporate defendant has been emptied. Between the award and enforcement, assets move. The response is the same as in domestic practice — move fast, take discovery where it is available, and pursue voidable transfers — but the cross-border version is slower and requires counsel in each jurisdiction.

Nobody checked whether the counterparty is an instrumentality of a state. Sovereign immunity from execution is discovered at the worst possible moment.

The arbitration agreement was signed by the wrong entity. In group structures, the contracting party is often a thinly capitalized subsidiary. GE Energy and the equitable estoppel doctrines it preserves may help, but the reliable fix is a parent guarantee that itself contains an arbitration clause.

Why the system holds

It is worth ending on the reason all of this works, because it explains why the defenses are read as narrowly as they are.

The Convention's bargain is reciprocal. Every contracting state agrees to enforce awards made in other contracting states, subject to a short and mostly procedural list of defenses, in exchange for its own awards being enforced elsewhere. A state whose courts read the public policy defense broadly, or that treats every procedural complaint as a due process violation, degrades the value of its own awards abroad. The incentive runs strongly toward narrow review, and after nearly seven decades the practice across most contracting states reflects it.

That is why an award travels better than a judgment. Not because arbitral tribunals are better than courts, but because 170-odd states made a mutual commitment that courts have honored — and because the commitment is self-enforcing in a way that comity between individual legal systems has never quite managed.

For a company doing business across borders, the practical lesson is that the arbitration clause is not boilerplate. It is the enforcement plan, drafted years before anyone needs it, and the ten minutes spent naming a seat and a set of rules is the highest-return ten minutes in the contract.