The Trojan Horse Rides Again: On the Swedish Supreme Court’s MFN Ruling in Okuashvili v. Georgia

On 26 June 2026, Sweden’s Supreme Court (“Court”) handed down judgment in Zaza Okuashvili v. Georgia, reversing the Svea Court of Appeal’s judgment and holding that an SCC tribunal could take jurisdiction over a claim against Georgia by combining the most-favored-nation (“MFN”) clause of one treaty with the dispute settlement clause of an entirely different one. The MFN clause’s capacity to reach into dispute settlement has often been described as a Trojan Horse, since contracting parties introduce the clause themselves. The Court has given the horse free rein by allowing an investor to switch arbitral forums via an MFN clause. This post makes two arguments: first, that the Court widened the MFN clause’s scope beyond the limits recognized by previous tribunals; second, that it failed to give due weight to the contextual requirement of State consent, overlooking considerations mandated by Articles 31 and 32 of the Vienna Convention on the Law of Treaties (“VCLT”). The facts of the case and the Court’s reasoning are set out in a previous Blog post in full (see here), this piece confines itself to two specific respects in which that reasoning falls short.

Dispute between Okuashvili and Georgia

Okuashvili holds both Georgian and British nationality (¶4). Due to Article 25(2)(a) of the ICSID Convention, Okuashvili was barred from arbitrating against his own State “regardless of what other citizenship he or she holds,” and this dual nationality excluded him from Article 8 of the UK-Georgia BIT that only offers ICSID arbitration (¶1, ¶5). The BIT’s MFN clause, Article 3, explicitly extends to Article 1 to 11 (¶2). Georgia’s separate treaty with the Belgo-Luxembourg Economic Union (“BLEU”) lets investors choose among ICSID, the ICC, or the SCC (¶3). Okuashvili used the MFN clause to reach into the BLEU treaty’s more permissive forum and commenced SCC arbitration against Georgia (¶6). The tribunal accepted jurisdiction; however, the Svea Court of Appeal set aside the award (¶7). The Court has now reinstated the tribunal’s approach (¶¶60-62). The intervening procedural history is set out in the previous Blog post referenced above.

Two findings of the Court are of importance. The Court held that ICSID and SCC arbitration are “of the same kind” under the ejusdem generis principle, since both are “established arbitration dispute resolution mechanisms” (¶46). It acknowledged that Article 8 reflects the “preferred” ICSID-exclusive version of the UK’s 1991 model investment treaty, deliberately chosen over a version offering investors a choice of fora – yet held that this choice “cannot be given any decisive importance” (¶47).

MFN and Dispute Settlement

Whether an MFN clause can reach a treaty’s dispute-settlement machinery has been debated for decades. The question was already argued in the 1950s, in the Ambatielos arbitration between Greece and the United Kingdom, where the ICJ confirmed the relevance of the ejusdem generis rule but dismissed the claim on its merits for failing to show any “privileges, favours or immunities… more extensive than those resulting from the basic treaty.”

It was Maffezini v. Spain that first used MFN as a weapon to create jurisdiction. The tribunal let an investor skip an eighteen-month local-litigation requirement under the Argentina-Spain BIT by importing a more permissive dispute-settlement clause from Spain’s treaty with Chile, reasoning that dispute settlement is “inextricably related” to investment protection and therefore falls within the clause’s reach (¶¶54-56). Siemens v. Argentina and Gas Natural v. Argentina followed suit, apparently settling the question in the investor’s favour.

However, other tribunals placed limitations on this rule. Plama v. Bulgaria and Salini v. Jordan held that an MFN clause cannot import dispute-settlement provisions “unless [it] leaves no doubt that the contracting parties intended to incorporate them” (Plama, ¶223). Wintershall v. Argentina’s stricter interpretation stated that an MFN clause cannot extend to procedural provisions without clear and unambiguous language. By 2015 the split had become pronounced enough that the International Law Commission’s Study Group devoted its entire final report to the MFN clause, concluding that tribunals would keep deciding the question case by case – not exactly the clear boundary needed. (¶216).

Two features of Maffezini’s own reasoning are directly relevant here and were overlooked by the Court entirely.

Extending the scope of MFN

The Maffezini tribunal, while expanding the MFN clause to dispute resolution, warned of the consequences and signaled that the clause’s scope should be narrowly construed. It held that the beneficiary of the clause should not be able to override fundamental conditions for acceptance of the agreement, especially if the beneficiary is the private investor (¶62). The tribunal listed the most prominent situations where the clause should not apply, including where “the agreement provides for a particular arbitration forum, such as ICSID, this option cannot be changed by invoking the clause, in order to refer the dispute to a different system of arbitration” (¶63). Okuashvili does swap ICSID for SCC arbitration, precisely what the Maffezini tribunal anticipated and excluded.

Plama and Salini confronted investors making this exact move and rejected it. In Plama, the tribunal refused to let a Bulgaria-Cyprus dispute travel from ad hoc arbitration into ICSID arbitration through the MFN clause, applying the “no doubt” standard quoted above (¶223). Salini reached the same conclusion on comparable facts (¶¶102–19). One case might support the Supreme Court’s conclusion: Garanti Koza v. Turkmenistan where the tribunal substituted an exclusive ICSID arbitration clause with another BIT’s dispute resolution clause that gave investors the option between ICSID and UNCITRAL. But that case drew heavy criticism for manufacturing consent the underlying treaty never gave.

None of this appears in the Swedish judgment. Instead, the Court works around the exceptions by finding that ICSID and SCC arbitration are “of the same kind” for ejusdem generis purposes, since both are established institutional forms of arbitration (¶46). This observation, while true, answers the wrong question. The forum-substitution exception has nothing to do with how institutions resemble each other; it has to do with whether a particular forum was the price of a State’s consent. Two arbitral institutions can be equally similar, and it will still be true that a State agreed to be sued in one of them and not the other. The extension of an MFN clause may be defensible where it removes an objectively burdensome procedural obligation that places an investor in a less favorable position than investors protected under comparable third-party treaties. The same justification, however, does not extend to investors seeking merely to substitute one arbitral forum for another based on their own interests.

Overlooking Interpretive Material

The second problem lies in how the Court read the treaty’s drafting history. The Court acknowledged that the UK offered two dispute-settlement templates in its 1991 model investment treaty: an ICSID exclusive “preferred” version, and a “alternative” version letting investors choose among ICSID, the ICC, and ad hoc arbitration (¶47). Georgia and the UK chose the exclusive one. A significant number of UK BITs concluded before 1991 used a single forum by default and the rationale behind introducing a broader template was to let treaty partners decide. Choosing between two available options is a form of expression, and arguably a more reliable one than the text of either option read in isolation, since it shows what the parties considered and rejected as well as what they accepted.

The Court rightfully identified that this question should be answered under Articles 31 and 32 of the VCLT. Article 31(1) of the VCLT treats this kind of surrounding practice as part of a treaty’s context. Article 32’s reference to the circumstances of a treaty’s conclusion points the same way. The Court weighs it for exactly one sentence, concluding that the UK’s alternative version “cannot be given any decisive importance in the interpretation,” without elaborating on why this is so.

The irony is that the Court had already committed itself to the opposite principle elsewhere in the same judgment invoking  effet utile - the idea that a treaty should be interpreted to avoid redundancy - to justify reading the MFN clause broadly (¶16). But effet utile cuts both ways. If the ICSID-exclusive template can be undone by a later MFN clause, then the “preferred” and “alternative” versions of the UK Model BIT lose their difference: whichever one a State signs, an investor can reach the other anyway. That is not giving effect to the treaty’s drafting choices.

Conclusion

This does not mean that MFN clauses should be shut out of dispute settlement. The holdings of previous tribunals, on procedural preconditions, remain sound. The objection here is narrower, and for that reason harder to wave away: on the specific question of substituting one arbitral forum for another, the Court has misapplied the one boundary that was set for that exact maneuver and treated a State’s considered choice between two negotiating templates as though it carried no interpretive weight whatsoever. The case now goes back to the Svea Court of Appeal on a narrow remand (¶61). Whatever that court does with it, other tribunals reading this judgment for guidance on the MFN question would do well to treat it as an outlier rather than a template.