Loading...
- Home
- Capabilities
- ACTIO® Hub
- About Us
- Connect with Us
- AP Library
June 3, 2026
Originally published in ACTIO E-Magazine Edition 29 (June 2026). Law as of June 2026.
Under Indonesian law, an AI system is not a legal subject, so it cannot bear legal responsibility itself. An automated contract is generally treated as a valid electronic contract if consent can be inferred and the four conditions of Article 1320 of the Indonesian Civil Code are met, and liability for loss caused by the system falls on a human legal subject, normally the operator, provider, or deploying party, assessed first through fault-based liability under Article 1365.
Indonesia’s contract law is being reshaped by agentic and automatic AI systems that can negotiate, draft, and execute agreements with minimal human intervention. At the same time, practitioners already report errors in AI-generated contracts and unresolved accountability gaps. Research cited in the source found that 38% of respondents use AI tools, mainly for legal drafting and research, while advocates in Semarang reported errors in AI-generated contracts and judges highlighted risks of bias and opacity if AI expands into case management without safeguards.
Article 1(17) of the ITE Law (Law No. 11 of 2008, as last amended by Law No. 1 of 2024) defines an electronic contract as an agreement between the parties made through an electronic system. Article 18 confirms that such transactions are binding, and Article 5(1) and (2) makes electronic information, electronic documents, and their printed results admissible as evidence, provided the agreement satisfies Article 1320. That article requires four conditions: consent, capacity to conclude an agreement, a specific subject, and an admissible cause.
Government Regulation No. 71 of 2019 (GR 71/2019) defines an “Electronic Agent” as a device of an Electronic System made to conduct an automatic action on certain Electronic Information which a Person conducts, and allows contracts to be concluded through an Electronic Agent that is operated. Article 47 of Government Regulation No. 80 of 2019 (GR 80/2019) expressly recognises that an Electronic Contract may arise from interaction with such devices and cannot be denied solely because it was generated automatically, unless the system is proven defective.
In practice, an automatic contract produced by an AI-driven contracting engine will generally be treated as a valid electronic contract if party assent can be inferred and Article 1320 is satisfied. Indonesian courts have accepted e-contracts in principle, but significant gaps remain around authentication, standard terms, and validity in more complex settings.
In the absence of an AI-specific statute, AI systems in Indonesia are best characterised as electronic agents or components of an electronic system rather than as independent legal subjects. That construction is reinforced by Circular Letter of the Minister of Communication and Informatics No. 9 of 2023 on Artificial Intelligence Ethics (MOCI Letter 9/2023), which defines artificial intelligence as a form of programming embedded in a computer system that enables the accurate processing and/or management of data. “Agent AI” is therefore an intelligent electronic agent whose decisions are attributable to an Electronic System Provider or other human legal subject, not a new juridical person capable of bearing rights and duties in its own name.
In the absence of a specific liability regime for AI, responsibility for harm may be assessed through existing civil liability principles, including fault-based liability where negligence, causation, and damage can be established. Article 1365 of the Indonesian Civil Code remains the central provision for extracontractual civil liability, requiring plaintiffs to prove an unlawful act, fault, loss, and a causal link. That traditional fault model runs into serious evidentiary problems, because AI autonomy and black-box opacity make it difficult for victims to trace programming choices, data defects, or oversight failures back to an identifiable human perpetrator.
In highly autonomous and opaque AI systems, insisting on proof of individual fault under Article 1365 creates a “legal impasse” that risks denial of justice. The source therefore proposes a shift towards strict or risk-based liability for AI operators and developers, and notes that Indonesian and comparative literature on AI in healthcare and finance recommends relaxing causation or burden-of-proof requirements where informational asymmetries between operators and victims are extreme.
Given that AI is framed as a tool or electronic agent rather than a subject, the source suggests analogising AI to things and subordinating it under Article 1367 (liability for persons and objects under one’s supervision) and Article 1368 (liability for animals). On this view, an AI-enabled contracting engine is more like a dangerous piece of equipment or a subordinate employee: if its actions cause loss, liability presumptively falls on the operator, owner, or employer unless they can show the loss was due to user misuse or force majeure. Strict liability may be considered for high-risk AI applications in future regulatory development.
The source sets out a four-step approach: map the AI stack onto the role hierarchy in Indonesia’s electronic transaction law and designate a single responsible operator with a recourse chain to the developer; allocate controllable risks through authority, evidence, and error-liability clauses, indemnities, and audit-trail logging, requiring human affirmation before an electronic agent expresses consent; ring-fence high-risk AI modules with human-in-the-loop, explainability, and audit obligations; and hard-wire good faith and professional ethics into engagement letters and vendor contracts so errors are transparently shared rather than silently absorbed by clients. The model operationalises Indonesia’s AI National Strategy 2020-2045 and mirrors tiered, risk-proportionate global frameworks such as the EU AI Act.
Can an AI system itself be sued for a faulty contract?
No. AI is not a legal subject under Indonesian law; liability must be attributed to a human legal subject, normally the operator, provider, or deploying party.
What makes an AI-generated contract valid?
It must satisfy Article 1320 (consent, capacity, a specific subject, and a lawful cause), and party assent must be inferable, for example through click-wrap, API triggers, or pre-agreed instructions.
Why is Article 1365 difficult to apply to AI?
Because it requires proof of an unlawful act, fault, loss, and causation, while AI opacity makes it hard to trace a defect to an identifiable human.
What is the proposed alternative?
A shift toward strict or risk-based liability for AI operators, including an analogy with Articles 1367 and 1368, so liability presumptively falls on the operator, owner, or employer unless the loss was due to user misuse or force majeure.
What should a contract say about AI use?
Name a single responsible operator, set authority, evidence, and error-liability clauses, require human affirmation before an electronic agent expresses consent, and require audit-trail logging.
As agentic AI moves from experimental pilots to the everyday engine of Indonesian contracting, the core challenge is institutional: whether civil law doctrine can keep pace without sacrificing legal certainty or fairness. A risk-sensitive mix of negligence, strict liability, and evidentiary presumptions, anchored in Articles 1320 and 1365-1367, would turn current accountability gaps into predictable, auditable duties for operators, platforms, and professionals.
