AI and Arbitration: The German Perspective

This is the fourth article in Clyde & Co’s latest international arbitration series covering Artificial Intelligence (AI) in international arbitration. In this piece, Counsel Florian Pötzlberger and Senior Associate Sophia Henrich from our Munich office provide a legal perspective from Germany.

I.    Setting the scene: AI, arbitration, and ongoing regulation

The European Parliament, Commission and Council reached a political agreement for the European Union Artificial Intelligence Act (AI Act) on 8 December 2023 and all 27 member states of the European Union (EU) endorsed the AI Act on 2 February 2024. Following a risk-based approach and impacting all industry branches, the AI Act establishes comprehensive rules and obligations for AI systems to manage their potential risks and ensure the public’s trust in AI technology. With a view on the rapid progress of technological development in the field of AI, the AI Act is groundbreaking and a major step forward in the regulation and governance of AI technologies. In addition, the EU is currently developing an AI Liability Directive as well as new regulations dealing with AI in the new EU Product Liability Directive. 

On 1 February 2024, the German Federal Ministry of Justice published a draft bill for the Modernization of Arbitration Law in Germany (Draft Bill). The proposals aim to further strengthen Germany's attractiveness as an arbitration location in international competition and to adapt Germany’s arbitration laws to the needs of today's world. In particular, the Draft Bill aims to align Germany’s laws to international developments, such as the revision to the UNCITRAL Model Law on International Commercial Arbitration in 2006, as well as revisions to the arbitration rules of other major arbitration institutions. Amongst other things, the Draft Bill includes new rules on freedom of form for arbitration agreements in commercial transactions and is aimed at strengthening transparency and promoting the digitalization of procedural law. 
 
In the light of those recent developments, we shed light on the role AI may play in the future of arbitration from a German perspective.

II.     AI in arbitration: a German perspective 

AI has a variety of potential use cases in arbitration, including document analysis, generating / editing text, evidence management, case prediction, decision-making, and drafting awards. Such uses serve to improve the efficiency and accuracy of arbitration proceedings and bolster the fairness of alternative dispute resolution (ADR). In this context, AI presents both challenges and opportunities for lawyers, arbitrators, institutions, and courts, potentially impacting their professional behaviour, autonomy, neutrality, confidentiality, and accountability. AI usage also has relevance under German law concerning the formal prerequisites, enforceability, and acknowledgment of arbitral awards. As in other professional areas, AI usage raises ethical and legal risks which must be considered.

We will provide you with a glimpse into the current developments and challenges of using AI in arbitration, as well as the opportunities and risks for all parties involved.

a.    Applicable law 

The legal framework governing the use of AI in ADR in Germany is complex and involves a myriad of sources and levels of law as a variety of legal issues may relate to any single AI usage (e.g. procedural law, confidentiality and data protection issues, transparency requirements, intellectual property, and professional conduct). 

Generally, the relevant legal framework includes applicable EU regulations (especially the AI Act, General Data Protection Regulation, AI Liability Directive, Product Liability Directive, Digital Single Market Directive, and Information Society Directive) as well as a variety of national laws (including the German Copyright Act, Unfair Competition Act, national data protection laws, professional regulations (BRAO, BORA), and criminal and civil laws). 

Particularly relevant for arbitration are sections 1025 et seqq. of the German Civil Procedural Code (ZPO) which recognises and regulates German arbitration principles, the respective arbitration rules of the arbitral institution chosen by the parties, and any further individual party agreements. Depending on the nature and context of a dispute, different rules and principles may apply and affect the validity, enforceability, or outcome of arbitration proceedings.

b.    Key principles: the use of AI in arbitration

The use of AI can have a substantial impact on various key principles in German and international arbitration. 

i.    Core principle: extended party autonomy

One of the main advantages of arbitration proceedings is greater party autonomy when compared to state court proceedings, as stipulated in Section 1042 ZPO. In light of the parties’ private autonomy and contractual freedom (sections 133, 157 of the German Civil Code) in determining the main features of arbitration proceedings, the parties are generally free to agree on AI usage. Nevertheless, the framework conditions of the ZPO are based on "traditional" arbitration proceedings, which may lead to limits in the recognition of arbitral awards, or the risk of an award being annulled in set-aside proceedings in cases where it is considered AI has been improperly used. 

The recently published Draft Bill, although drafted in the light of “the continuously advancing digitalisation of procedural law”¹  does not aim for a fundamental change in this respect and does not even contain any explicit regulation of the use of AI in arbitration. Rather, regarding technical innovations, it deals with comparatively minor adjustments which are already widely used in practice, such as the admissibility of (e-)awards contained in an electronic document and signed using a qualified electronic signature, and the conduct of oral hearings via video and audio transmission.

It is clear that the intention at this stage is that AI should not replace traditional arbitration proceedings or the arbitral tribunal but should be understood as a supportive tool to be used in the parties’ interests.

ii.    Basic procedural guarantees

The use of AI by the parties to an arbitration may involve a variety of procedural implications relating to, amongst other things, due process, confidentiality, data protection, and transparency issues. In any event, the use of AI in arbitration must not lead to a violation of the minimum standards of fair proceedings and the German ordre public to ensure the enforceability of the award, c.f. section 1059 (2) no. 2 lit. b ZPO.

However, these principles place only a few limits on party autonomy. In particular, the concept of a fair proceeding precludes agreements by which the parties waive in advance the requirement of objective arbitrability, the requirement of equal treatment under procedural law, and the right of each party to have the opportunity to present their case in arbitration proceedings². Accordingly, the parties cannot agree on technical applications that fail to uphold these principles (e.g. due to biased or hallucinating AI applications). In this event, the arbitration award could be annulled at a party’s request in set-aside proceedings, irrespective of a prior agreement between the parties (section 1059 (2) no. 1 lit. d ZPO). The Draft Bill does not touch on these basic principles. 

c.    AI usage by the parties to an arbitration

The parties to an arbitration may use AI, for example, to transfer, view and analyse large volumes of data, and to support the creation of standardized documents (document automation / generation), including automated document evaluation and the semi-automated creation of drafts³. AI tools may also be used for identifying potential arbitrators and creating schedules and structures for arbitration proceedings. Document disclosure and submission are further fields where AI already is or may be a useful supporting tool for strengthening the efficiency of arbitration proceedings, especially in the international context. 

The parties and their arbitration counsels, however, need to consider all the legal risks and challenges connected with the use of AI, including but not limited to a careful review, as well as amendments and / or changes to all AI-generated content. Of course, all professional and contractual confidentiality obligations need to be fulfilled, for example, by research and AI usage based on anonymized data and / or services by AI operators providing secure systems for certain (supervised) professions. It is imperative that third party rights, including intellectual property, are respected at all times.

d.    Decision making by the arbitral tribunal 

The possibilities of AI not only impact on the parties' room for manoeuvre, but also offer new opportunities for appointed arbitrators.

The core of an arbitrator’s duty is undoubtedly decision-making and drafting of decisions in the form of binding arbitration awards. Considering the capabilities of AI already available, it is reasonable to believe that AI applications could in the future be trained to conduct an assessment based on the applicable law and the facts of a case, thus providing support in the decision-making process. From a technical perspective, decision-making carried out entirely by AI is by no means an impossible idea. However, and while the parties to arbitral proceedings are generally free to choose their arbitrators (section 1034 et seq. ZPO), in order for the arbitration award to be recognized in Germany beyond any legal doubt and to avoid the risk of annulment in set-aside proceedings (sections 1060 (2), 1059 (2) ZPO), the essential elements of the decision-making process must remain a human responsibility. 

Article 92 of the German Grundgesetz (the German Basic Law) provides that the judicial power and core functions of the judiciary are entrusted (“anvertraut”) to judges as individuals. The prevailing view in legal literature reflects this sentiment, providing that the concepts underlying the current sections 1025 et seqq. ZPO⁴, as well as the proposed amendments in the Draft Bill, presume a human arbitrator. Indeed, the same is true for the New York Convention on the Recognition and Enforcement of Foreign Arbitral Awards⁵.

While the duty to reach a decision in compliance with procedural law and the guarantees of the rule of law remain a highly personal duty of any appointed arbitrator⁶, this premise leaves room for the utilization of AI as support for human decision-making. Given that both organizational and content-related auxiliary activities (e.g. preparing material, creating overviews, timelines or schedules, and legal research) can be carried out by assistants without the recognition of the arbitral award in Germany being excluded in view of sections 1025 et seqq. ZPO, the usage of corresponding AI applications should also be possible to the same extent. 

In consequence, utilizing AI for such auxiliary activities (e.g. to generate a procedural timeline or a verbatim record of an oral hearing) should not give rise to significant legal risk, at least insofar as the generally applicable requirements for the use of technical systems, including data protection laws and the future AI Act, are complied with. Beyond this, the use of AI for activities within the scope of the arbitrator's duties depends mainly on the consent of the parties. If the parties agree to the specific use of AI, the fact that AI is involved cannot in itself be a ground for annulment of the respective decision and does not automatically render an arbitral award non-binding⁷.

An arbitrator using AI will nevertheless have to observe the existing general requirements for the utilization of AI, in particular data protection, as well as any agreements between the parties. Under the current draft of the AI Act⁸, systems that are intended to support judicial authorities in the investigation and interpretation of facts and legal provisions, and in the application of the law to specific facts, are classified as high-risk and thus subject to stricter requirements. 

The current draft of the AI Act further stipulates the same classification to apply to AI systems intended to be used by ADR bodies for such purposes when the outcome of the ADR proceedings produce legal implications on the parties, as is the case in arbitral proceedings. Systems which are intended for purely administrative support activities and do not affect the actual administration of justice in individual cases are excepted. Also, the AI Act recognizes that the final decision-making shall remain human-driven. 

Below this threshold, the use of AI is likely to benefit procedural efficiency in arbitration in many respects in the future.

III.    What the future may bring

To sum up, arbitration in Germany and around the globe will not lose its scope of application in the foreseeable future due to AI technologies. Human judgment and negotiation strategies are too important in resolving disputes efficiently and economically in international arbitration. However, the use of AI will simplify various processes and make them more efficient, which will generally benefit all parties to arbitration. Concurrently, the legal risks and challenges outlined above must be thoroughly considered at all times. It remains to be seen which AI use cases will be predominant in international arbitration. Considering the rapid technological development in the field of AI, further developments will remain thrilling from a legal and technological perspective.

This series will continue next week with the perspective from Poland.

¹ German Federal Ministry of Justice, Draft Bill for the Modernisation of Arbitration law (English translation of the German draft), available at: Draft Bill for the Modernisation of Arbitration Law (bmj.de) (accessed 12 February 2024).

² Kreis, in: Kaulartz/Braegelmann, Rechtshandbuch Artificial Intelligence und Machine Learning, 1st edition 2020, Chapter 14.1, para. 41.

³ Rühl, in: Kaulartz/Braegelmann, Rechtshandbuch Artificial Intelligence und Machine Learning, 1st edition 2020, Chapter 14.1, II. 1. para.4 et seq.; Dörrscheidt/Finkelnburg, Digitale Streitbeilegung in Deutschland, LTZ 2023, 255.

⁴ Münch, in: Münchener Kommentar zur ZPO, 6th edition 2022, Vorbemerkung zu § 1025 ZPO, para. 5 with references to Article 92 Basic Law; Dörrscheidt/Finkelnburg, LTZ 2023, 255 (259).

⁵ In detail: Scherer/Jensen, Die Digitalisierung der Schiedsgerichtsbarkeit, in: Riehm/Dörr (eds.), Digitalisierung und Zivilverfahren, 1st edition 2023, Chapter 25, 614.

⁶ Münch, in: Münchener Kommentar zur ZPO, 6th edition 2022, Vorbemerkung zu § 1025 ZPO, para. 5; Dörrscheidt/Finkelnburg, LTZ 2023, 255 (259); Scherer/Jensen, in: Riehm/Dörr (eds.), Digitalisierung und Zivilverfahren, 1st edition 2023, Chapter 25, 617.

⁷ Kreis, in: Kaulartz/Braegelmann, Rechtshandbuch Artificial Intelligence und Machine Learning, 1st edition 2020, Chapter 14.1, para. 50.

⁸ For an overview please see AI Act leaked : Clyde & Co (clydeco.com).