May 15, 2025
Originally published in ACTIO E-Magazine Edition 26 (June 2025). Law as of June 2025.
Artificial intelligence is transforming dispute resolution, but it cannot yet serve as an arbitrator under Indonesian law. Law No. 30 of 1999 requires arbitrators to be persons, and AI is neither a natural person nor a legal entity, so it cannot meet the statutory requirements. An arbitration agreement appointing an AI arbitrator risks being invalid, and any resulting award may be unenforceable in Indonesia. AI should assist human arbitrators, not replace them.
The utilisation of Artificial Intelligence (AI) has increased rapidly over the years. In the legal field, this is clear from the growing use of AI-based tools such as Solomon and Lex Machina, which read court decisions, predict outcomes, and suggest resolution strategies.
Against that background, a new, potentially controversial question has arisen: can AI serve as an arbitrator? In Indonesia, there is no specific legal basis for the use of AI arbitrators. It could therefore be interpreted as not yet permitted under Indonesian arbitration law. The answer matters, because appointing an AI arbitrator may invalidate the arbitration agreement and make any resulting award unenforceable.
Under Article 1(1) of Law No. 30 of 1999 on Arbitration and Alternative Dispute Resolution, arbitration is a way of settling civil disputes outside the general court based on a written arbitration agreement between the disputing parties. Arbitrators are defined in Article 1(7) as one or more persons chosen by the disputing parties or appointed by the District Court or an arbitration institution to render a decision on a particular dispute.
The reference to “persons” is decisive. Under Indonesian law, the legal subjects are natural persons and legal entities. AI is neither a natural person nor a legal entity, so it cannot count as an arbitrator as currently defined.
Although Indonesian law does not expressly prohibit the appointment of an AI arbitrator, an AI arbitrator would also need to satisfy the requirements of Article 12(1) of the Arbitration Law. Any party eligible to be appointed as an arbitrator must have full capacity to perform legal actions and must have mastered and understood the dispute, as required under Article 12(1), which states:
“Unless otherwise agreed by the parties, the arbitrator or the Arbitral Tribunal appointed shall fulfil the following requirements: (a) is capable of carrying out legal acts; (b) is at least 35 years of age; (c) has no family relationship by blood or marriage up to the second degree with either of the opposing parties; (d) has no financial or other interest in the award; (e) has a minimum of 15 years active experience in his field.”
An AI will not generally be considered to have the legal capacity to perform legal acts. Age is also a matter of interpretation: in the case of an AI, the age of the software or the technology could be relevant. Further, given that AI relies on existing decisions and data trained by its programmer, it may carry biases, which creates a risk of violating the impartiality requirement.
The parties’ freedom of contract offers another route. Article 1338 of the Indonesian Civil Code incorporates the principle of freedom of contract: generally, the parties may agree on any terms, provided they do not contradict the law, public order or good morals. In theory this principle could open the door to AI arbitrators. Nevertheless, the absence of specific regulations leads authors to conclude that the use of AI arbitrators is not yet fully feasible or valid.
If AI arbitrators had been appointed under an arbitration agreement, the likelihood of enforcement failing increases significantly. This is generally caused by the failure to fulfil the requirements for a valid arbitration agreement, particularly the absence of legal capacity as defined in Article 1320 of the Indonesian Civil Code.
In Indonesia, the arbitration agreement is essential because it assigns the parties’ dispute to arbitration and establishes the arbitrators. If an arbitration agreement with an AI as the arbitrator is invalid, then actions and awards under that agreement are also invalid.
Related to the enforcement of foreign arbitral awards rendered by AI arbitrators, the 1958 New York Convention would pose challenges. Article V(1)(a) of the New York Convention provides that if the agreement was invalid under the law to which the parties have subjected it, an arbitral award may be refused. Furthermore, Article V(2)(b) states that the court where enforcement is sought may refuse recognition and enforcement if it considers such recognition contrary to the public policy of that country.
The authority responsible for determining the recognition and enforcement of foreign awards is the court, which must first verify the legitimacy of the arbitration agreement. Given that an AI arbitrator does not meet current legal requirements, the court may reject the validity of the arbitration agreement and refuse enforcement.
Even where the parties have consented to the appointment of AI arbitrators, issues may arise as regards the validity of the arbitration agreement under Indonesian law, primarily due to concerns about legal capacity requirements.
Despite the absence of a ban on AI, European Union Member States have issued the EU AI Act, which provides guidelines for the development and utilisation of AI in legal decision-making, including arbitration. The Act was enacted on 1 August 2024 and will be fully enforced on 2 August 2027. It protects rights while encouraging technological advancement. It establishes a risk-based framework that classifies AI systems according to their level of risk, ensuring that AI development is done responsibly and ethically.
In Indonesia, the PDP Law and ITE Law offer a foundation for regulating AI and personal data processing. However, Indonesia has not yet developed specific regulations governing the use of AI in dispute resolution. Legislators have made progress by drafting the Artificial Intelligence Technology Bill, but it remains in the planning stage and is expected to be submitted to the DPR in 2025. The bill adopts a risk-based framework to set rules for the development and utilisation of AI in Indonesia. Its aim is to ensure that personal data is adequately protected and used responsibly.
For now, AI is best seen as an assistant to arbitrators, not as an arbitrator itself. Its role in arbitration should focus on process support and decision-making assistance, with human arbitrators retaining the authority to decide. AI could also be used as a foundation for building other legal technologies, such as predictive analytics, that provide strategic insights.
Nonetheless, AI technology should be adopted cautiously. It can provide useful interpretations of how arbitration panels are likely to decide, but parties should be aware of its limitations and proper use. For example, AI cannot yet understand nuance, and its predictive results are based on patterns from past decisions; disputes resolved by arbitral tribunals are, however, fact-specific. If legal professionals rely too heavily on AI, they risk losing critical thinking skills and eroding trust in the final decision.
Several arguments weigh against the use of AI arbitrators. First is the question of emotional intelligence and human judgment. Awards rendered by arbitrators may be perceived as more convincing or legitimate if they are produced by a human, rather than a machine. Some parties may prefer human arbitrators because they can grasp the subtle aspects of all the facts and arguments of a case. AI does not yet fully measure up in this regard.
Furthermore, AI arbitrators may face issues regarding the confidentiality of information, the security of sensitive legal documents and dependence on AI by human arbitrators. Assuming that AI technology could be used as an arbitrator, there would also be data privacy concerns if AI is linked to online platforms or servers.
Also, AI may not even produce accurate results. AI systems such as ChatGPT have been found to be prone to hallucinations, in which they provide incorrect or false information. This is a persistent limitation in current AI models, and it still cannot be fully remedied. AI may also produce biased decisions if it was trained using biased training data, which would compromise the impartiality and objectivity of its decisions.
While digital assistants can be helpful, full delegation to AI does not appear to be equitable. The only known case of an AI arbitrator is the AI Arbitral Tribunal, created at the VIII BRICS Legal Conference held in Moscow on 22 August 2024. There, an AI arbitrator rendered an award on a real dispute, having analysed procedural documents, the submitted facts and arguments, the parties’ positions and other pertinent evidence. The award was accepted by the parties and enforced under the terms of a mediation agreement. However, AI’s use as an arbitrator remains an unprecedented experiment, and the oversight of human intervention probably remains subject to debate.
The Chartered Institute of Arbitrators (CIArb) published its Guideline on the Use of AI in Arbitration on 14 August 2025. It addresses the use of AI tools in arbitration, including their ethical use. Although the guideline does not cover the appointment of AI as an arbitrator, it emphasises when AI may be used in arbitration, the disclosure requirements, human oversight, and the validation of outputs — most of which involve the use of AI by arbitrators, not AI as the arbitrator.
When it comes to foreign awards rendered by AI arbitrators, the enforcement problem remains. The lack of specific regulations on AI arbitrators may mean that parties who have agreed to an AI arbitrator could find themselves unable to enforce an award in Indonesia.
No. Under Law No. 30 of 1999, an arbitrator must be a “person” — a natural person or a legal entity. AI is neither, and cannot meet the statutory requirements of Article 12(1), including legal capacity and a minimum age of 35.
Not expressly. Indonesian law does not expressly prohibit AI arbitrators, but no law permits them either, and an AI arbitrator would fail the statutory eligibility requirements.
It may open the door. Under Article 1338 of the Civil Code, parties may generally agree on any terms, provided they do not contradict the law, public order or good morals. But an AI arbitrator’s lack of legal capacity still makes the agreement vulnerable to invalidity.
Unlikely. An AI-rendered award risks refusal under Article V(1)(a) of the New York Convention if the arbitration agreement was invalid, and under Article V(2)(b) if enforcement would be contrary to Indonesian public policy.
AI should assist human arbitrators — reading decisions, predicting outcomes and suggesting strategies — while humans retain the authority to decide.
Indonesia’s arbitration law awaits modernisation. Until the law establishes a clear framework for AI arbitrators, including issues of legal capacity, impartiality and the validity of AI-rendered awards, the use of AI as an arbitrator appears neither feasible nor valid under Indonesian law. AI technology will only be fully effective if the law and regulations keep pace with technological developments. For now, AI is best positioned as a support tool for human arbitrators, rather than as an arbitrator itself.


