2025 in Review: Arbitration-Related Developments in France
February 2, 2026
2025 was a year of transition for France being in the midst of reforming its arbitration law. This post highlights significant developments and judicial decisions of the past year.
Reform of French Arbitration Law: Progress Amid Turmoil
The reform process launched in 2024 (see previous coverage) continued in 2025, albeit with more turmoil than recent reforms in the UK or Germany. The Working Group established by the Ministry of Justice issued its proposals in March 2025, eliciting mixed reactions among institutions, with some expressing support (AFA, CMAP or Paris Place d’Arbitrage) alongside others adopting a more cautious stance (CFA, ICC France). Despite legitimate concerns, the co-chair of the Working Group addressed criticisms in a manner atypical in academic circles (see here).
The first step announced by the Minister of Justice in April 2025 is underway with the adoption of a decree “tightened around largely consensual proposals” (18 out of 40 proposals). A draft decree was circulated, which stakeholders can comment on before 20 January 2026, suggesting that earlier concerns over lack of consultations have been partly heard.
The draft includes a set of proposals codifying praetorian solutions, such as the definition of an “arbitral award” which restates the French courts’ well-established definition (see e.g., this year, Mesa and Oschadbank) and the amended definition of “international arbitration” referring to “international economic” interests, to reflect the evolution of the courts’ assessment (see e.g., this year, iXblue).
It also includes enhancements, including an ad-hoc-friendly provision inspired by Article 9 of the ICC Rules allowing (subject to all parties’ consent) multi-contract claims under one or more arbitration agreements to be heard within a single arbitration, the recognition of electronic awards, the tribunal's power to liquidate penalties, and the removal of the suspensive effect of set aside proceedings in domestic arbitration.
The draft however also includes more controversial points, such as the removal of form requirements for arbitration agreements – which does not carry much interest in practice and could spur superfluous litigation as to the existence and scope of these agreements (see Keppel) – and the supporting judge's power to grant enforceability to decisions that are not subject to exequatur (e.g. provisional measures), which could dissuade investment in France.
Investment Arbitration: Tightening Jurisdictional Review
For several years, the review of investment arbitration awards, where jurisdiction plays a pivotal role, has allowed French courts to refine the scope of their review of jurisdiction under Article 1520(1) of the Civil Code of Procedure (“CCP”). 2025 was no exception.
Albeit not limited to investment arbitration, the first decision to mention is the major (and welcome) shift decided by the Paris Court of Appeal (the “CoA”) in Astaris, in which it departed from the so-called Schooner approach (reported in the 2020 year-in-review) in saying that the waiver under Article 1466 CCP fully applies to Article 1520(1) CCP and precludes investors from presenting new jurisdictional arguments at the annulment stage. In doing so, the CoA emphasized the importance of procedural loyalty and anticipated one of the Working Group’s reform proposals. It should be hoped that the French Supreme Court ("Cour de Cassation") confirms the shift.
Another important decision is the Cour de Cassation’s ruling in Agarwal, in which it quashed the CoA and recalled that the scope of review of the annulment judge does not depend on the qualification made by the arbitral tribunal. If the latter has wrongly declared itself incompetent on merits grounds (here, because the investor did not hold the investment at the time of the breach), the annulment judge is powerless to set aside the award because this would amount to reviewing the merits of the case.
To counterbalance their in-depth scrutiny power under Article 1520(1) CCP, French courts have developed over the years case law limiting their scope of review to issues directly related to the concepts of investment and investor, as defined by the underlying BIT and its dispute settlement clause. In 2025, French courts refused to review, for example, the legality requirement (Cengiz and Nurol), the scope ratione temporis of a BIT when it involves the interpretation of the applicable law (Selmani), the date of the investment (Oschadbank), the ownership of the investment at the time of the breach (Agarwal), compliance with the BIT’s requirement for prior recourse to domestic courts (Agarwal), and the applicability of a BIT fork-in-the-road clause (Olin). A contrario, French courts accepted to review, for instance, the date of the dispute (Oschadbank and Nurol), and whether the investment was made in the territory of the host State (Oschadbank).
A notable decision was also rendered in Hydro, in which the CoA declined Albania’s annulment request against an ICSID award as well as the decision of an ICSID annulment committee that upheld it. The CoA logically grounded its decision on Article 53(1) of the ICSID Convention which expressly bars court intervention and noted that Albania had subscribed to ICSID’s annulment system.
Independence and Impartiality: A Persistent Blind Spot
While 2024 proved that arbitrators’ independence and impartiality remain a critical concern (rare annulments in Telitalia and CSF, also confirmed in the PAD case), 2025 unfortunately showed little progress, as this paramount issue continues to be poorly addressed.
In CWT, the arbitrator failed to disclose family ties with the director of a company listed as an “other entity concerned” in the ICC “case information sheet”. The Cour de Cassation deemed this listing declaratory in nature and that it did not evidence the company’s interest in the arbitration. On this basis, annulment was denied. While factually sound (the company had no interest in the outcome of the arbitration), the decision perpetuates uncertainty around disclosure duties and invites arbitrators to silence.
Likewise, it is disappointing that the reform does not address the issue, which suggests that disclosure duties under French law is likely to remain a blind spot in the future.
Other rulings clarified that the amicus curiae intervention of an arbitrator, five years after the arbitration, in an unrelated case involving the respondent State, on a totally unrelated issue, does not warrant annulment (Oschadbank). The same conclusion was reached in Koweit, in which five repeated appointments by the same law firm were deemed old (6, 5 and 2 years) and well-known, while the most recent one (a few months before) was not problematic because there had been a change of counsel.
On a positive note, the CoA found in Koweit that independence and impartiality can only be raised under one annulment ground (irregular constitution of the tribunal – Article 1520(2) CCP) and can be waived under Article 1466 CCP, whereas it had previously held in Pharaon (reported in the 2021 year in review) that it could also be raised with new arguments under Article 1520(5) CCP (public policy).
Turning Litigation Clauses into Arbitration Agreements: Lessons from Keppel
In Keppel, the CoA dismissed a written litigation clause in favour of an oral arbitration clause. Keppel responded to a call for tenders whose general conditions included a litigation clause referring to Qatari courts. In its tender application, Keppel requested the litigation clause to be replaced by an arbitration agreement. Such request was repeatedly made during the negotiations and had been discussed by the parties, who formally agreed to a mediation phase. Seized of a dispute between Keppel and the contractor, the arbitral tribunal declined jurisdiction.
To infer the parties’ common intent, the CoA relied on , among other things, (i) the fact that the arbitration agreement was an essential condition of Keppel’s tender application, (ii) the parties’ agreement to mediation incompatible with the litigation clause, and (iii) the contractor’s agreement with the documents discussed the day before the attribution of the tender to Keppel. Interestingly, the CoA considered that the “priority of documents” clause does not supersede the parties’ common intent.
In doing so, the CoA elevated the parties’ conduct and genuine consent over the literal wording of the contract. The ruling carries practical lessons: correspondence and oral exchanges can be as important as contract drafting, and the choice of Paris as a seat may allow a litigation clause to be transformed into an arbitration agreement through the parties’ behavior.
Corruption Allegations: Towards a More Restrictive Approach
While in the past years French courts have annulled awards based on vague indicators of corruption (see, e.g., Sorelec, reported in the 2020 year in review), it seems that the courts are now more inclined to give effect to the tribunals’ assessment of corruption and to safeguard the award.
In Averda, the CoA scrutinized an award which had recognized corruption in relation to one contract, while upholding claims for payment under other contracts that were clear from corruption. The Court first noted that annulment is warranted only if recognition and enforcement in France enables a party to benefit from corruption. Relying on the “thorough instruction” of the tribunal, the CoA observed that (i) only a portion of services under one contract was affected by corruption, and (ii) the corrupted practices were neutralized through an abatement rate on the sums owned, never challenged by the applicant.
The decision shows that evidence of corruption does not necessarily lead to annulment. Corruption must not only exist but also produce effects in France. It further reflects that the CoA’s in-depth review under Article 1520(5) CCP does not prevent it from relying exclusively on the tribunal’s reasons when corruption allegations have been rigorously addressed. It finally underscores that the parties’ conduct plays a role in the assessment of corruption.
Stay of Enforcement: A Narrow Path with Room for State Interests
Traditionally, French courts have strictly applied Article 1526 CCP and its requirement that stay of enforcement requires that it is “likely to seriously infringe the rights” of a party (see previous coverage). This restrictive approach did not changed in 2025 (see e.g., Eova, H2 Holding, Alessi Domenico, Hilton), but two decisions granting stay to States are worth mentioning.
In OMV, a Romanian Ministry sought a stay of enforcement pending a decision of the European Commission (“EC”), arguing that the adverse award amounted to state aid prohibited under EU law. While it had initially granted the request, the CoA ultimately overturned it after observing that the Ministry had not formally notified the EC two years after the award was rendered, nor had the EC launched an investigation despite being aware of the award. The initial granting of the stay suggests that the CoA is aligned with the Court of Justice of the European Union ruling in Micula and that States may increasingly leverage this approach to temporarily stall enforcement of adverse awards.
In Enka, the CoA granted Georgia’s stay request on the grounds that enforcement would have a disproportionate impact – the EUR 400 million adverse award represented “nearly 60% of the defense budget” and “around 250% of the justice budget” of the State, thus carrying out a serious risk of harm to the State.
Sulu Saga: Closing Chapter?
While another procedure in the Sulu saga (see previous coverage here and here) was summarily dismissed by an ICSID tribunal on the ground that neither the expenditures on legal fees nor monetary interest in the ad hoc arbitration award constituted a BIT-protected investment, the Cour de Cassation confirmed the definitive annulment (reported in the 2023 year in review) of the multibillion award rendered in the ad hoc arbitration. It recalls how overly rigid adjudicator designation (without fallback mechanism) can undermine the effectiveness of an arbitration agreement.
You may also like