Is AI a Person, Property, or a Legal Loophole in Indonesia?
Imagine this scenario: You use an AI-powered financial advisor app to manage your investments. The AI hallucinates, gives you disastrous advice, and you lose millions of rupiah. Or perhaps, an AI chatbot deployed by an Indonesian e-commerce site promises a massive discount to a customer by mistake.
When things go wrong, a fundamental legal question arises: Who do you sue? Can you drag the AI to court? Do you sue the software developer? Or does the blame fall on the company using the AI? We human love the blame game, right? Some say we have been playing this game since the beginning of time when Adam blamed Eve and Eve blamed the snake. I hardly play it. But for practical reasons, let’s play it for a sec.
As artificial intelligence becomes deeply integrated into Indonesia’s daily life and business landscape, our legal system is being put to the test. To understand who is responsible for AI's actions, we first need to answer how Indonesian law views AI. Is it a person, a piece of property, or something else entirely?
The first question that we need to answer is AI a human or a property, or electronic agent? To determine liability, the Indonesian legal system (rooted in the Civil Code or KUHPerdata) divides the world into two main categories: legal subjects (subjek hukum) and legal objects (objek Hukum).
Is AI a legal subject (A Person)? In Indonesia, only humans (natuurlijke persoon) and recognized legal entities like limited liability companies or PTs (rechtspersoon) are considered legal subjects. They have rights, can own property, and most importantly, can be held liable in court. The verdict? No matter how "smart" or human-like ChatGPT or Claude seems, AI is not a person under Indonesian law. You cannot sue an algorithm, and an AI cannot pay you damages or go to jail.
Is AI a legal object (property)? Legal objects are things that can be owned and controlled by legal subjects. Traditionally, software and algorithms fall under this category as intangible assets and intellectual property (specifically protected under the Indonesian Copyright Law). However, calling modern AI mere "property"—like a chair or a static piece of code—fails to capture its autonomous nature. A chair doesn't generate new images or make automated decisions.
The sweet spot: AI as an "Electronic Agent". This is where Indonesia’s Information and Electronic Transactions Law (UU ITE) steps in to save the day. Under the ITE Law (most recently amended by Law No. 1 of 2024), AI fits perfectly into the definition of an Electronic Agent (Agen Elektronik).
Article 1 Point 8 of the ITE Law defines an electronic agent as a device of an electronic system created to perform actions regarding certain electronic information automatically, which is operated by a person. By legally classifying AI as an electronic agent, Indonesian law acknowledges that AI is a tool capable of independent, automated action, but it remains fundamentally tied to a human or corporate operator.
The million-dollar question is who is responsible for AI output? Because AI is legally an "electronic agent" and not a person, it operates under a simple legal reality: The tool cannot be blamed; the operator must take the fall.
Article 21 of the ITE Law explicitly states that every electronic agent operator (penyelenggara agen elektronik) is legally responsible for the operation of the electronic agent they manage. But in the complex world of AI, who exactly is the "operator"? liability usually trickles down to three parties:
The first party is The Developer (the creator of the AI). If an AI model has an inherent flaw, bias, or security vulnerability from its base code, the original developer (e.g., OpenAI, Google, or a local Indonesian tech company) could be held liable. However, most AI developers protect themselves with ironclad Terms of Service, stating their models are provided "as-is," shifting the risk to the companies that use them.
The second party is The Deployer (the company Using the AI). If your Indonesian startup integrates an AI chatbot into your customer service, you become the Electronic Agent Operator in the eyes of your customers. If your AI defames a competitor, leaks a customer's personal data (a strict violation of Indonesia's Personal Data Protection Law / UU PDP), or causes financial harm, Article 1365 of the Indonesian Civil Code (Perbuatan Melawan Hukum / Tort) combined with the ITE Law dictates that your company must pay the compensation. You cannot tell the judge, "The AI went rogue!" Under Indonesian law, you deployed it; you own the risk.
The third party is The User (the person prompting the AI). Liability can also fall on the end-user. If a user intentionally feeds malicious prompts into an AI to generate deepfakes, copyright-infringing material, or defamatory content (pencemaran nama baik), the user is the one manipulating the Electronic Agent. In this case, the ITE Law will target the user who pressed "enter."
The "AI defense" does not exist in Indonesian courts. If your business is adopting AI to cut costs or improve efficiency, you must treat it like hiring a highly capable but reckless employee. To protect your business under the ITE Law and the Civil Code you must take several measures as follow: Audit Your AI to ensure the AI tools you use do not scrape or expose private consumer data without consent. You should also update your terms and conditions and explicitly state the limitations of your AI's advice (especially in finance, health, or legal sectors). And the most important measures to be done is to keep human oversight on the AI as the ITE Law requires Electronic Agents to have features that allow users to correct errors. Always have a "human in the loop" (a human customer service override) to prevent an AI's mistake from turning into a corporate lawsuit.
AI is no longer just a futuristic concept; it is a present-day legal liability. In the eyes of Indonesian law, behind every artificial intelligence, there is a real-world entity holding the remote—and the legal system knows exactly where to look when things go wrong.