AI is a tool, not a legal subject and not a substitute for an advocate’s professional judgment. The advocate who uses AI carries full responsibility for whatever legal consequence follows.
Technology has changed how advocates work. Legal research that once took days can now be finished in minutes. Contract drafting, judgment summaries, and cross-language translation have started moving through generative AI (“Gen AI”) in Indonesian law firms.
Gen AI use in the legal sector has spread into work that used to be entirely manual, from legal research and document review to contract drafting and opinion preparation. The technology lets advocates produce legal documents faster.
Legal work still holds something that cannot simply be handed to a machine, though. Client confidentiality, professional judgment, and accountability for every word submitted to a court remain with the person, not the tool.
On 6 August 2026, the National Board of the Indonesian Advocates Association (Perhimpunan Advokat Indonesia, “PERADI”) launched its Guidance on the Use of Artificial Intelligence for Indonesian Advocates. PERADI’s Standing Committee on Legal Technology and Innovation prepared the guidance by drawing on the American Bar Association, the State Bar of California, the Council of Bars and Law Societies of Europe, the Law Society of New South Wales, the Law Institute of Victoria, and Singapore’s Ministry of Law, then aligning those references with Indonesia’s Advocates Law and Personal Data Protection Law.
The question that now matters for advocates and law firms is no longer whether AI use is permitted. It is what changed in their professional obligations once the guidance took effect.
AI as an Assistive Tool, Not a Substitute for an Advocate’s Judgment
PERADI’s Chairman, Ahmad Fikri Assegaf, set out where Gen AI sits in the guidance. In his words, AI is positioned as an assistant, not a replacement for human reasoning. He has also warned of “cognitive debt,” the decline in analytical ability that comes from leaning on machine output too heavily.
That principle sits behind the guidance’s “human in control” standard. Gen AI can spark an initial idea, but the final decision and the responsibility for it stay with the advocate. Robert F. Sidauruk, who coordinates PERADI’s Standing Committee on Legal Technology and Innovation, has said the guidance isn’t meant to restrict AI use or frighten advocates, but to give them a clear footing.
The consequence plays out at the level of daily practice. Gen AI output is probabilistic, not deterministic. The model predicts the statistically likeliest next words rather than weighing legal principle or a client’s best interest. The guidance frames this as a conceptual limitation, meaning fluent-sounding output can still be wrong, incomplete, or entirely invented.
Because of that limitation, Gen AI carries no professional responsibility and answers to no code of ethics. Accountability stays with the advocate as a human being, regardless of how much of the work AI touched.
Seven Duties That Stay With the Advocate
PERADI’s guidance turns that principle into concrete obligations spread across different areas of practice.
Maintaining Confidentiality of Client Information
Entering confidential client information into a public Gen AI tool that uses user input to train its model is equivalent to disclosing that information to a third party. Before processing a document, advocates must scrub or redact party names, addresses, transaction values, and identification numbers, though the guidance is clear that this step is an additional safeguard, not a substitute for choosing a genuinely confidential tool in the first place.
Verifying AI-Generated Information
Legal research is where hallucination risk runs highest, since output can include case citations or regulatory references that look authentic but are fabricated. Advocates must independently verify every AI research result against official sources. The guidance flags one error to avoid, using one Gen AI tool to verify another’s output, since a model can confidently affirm the existence of a case that was never decided.
Preserving Independent Professional Judgment
Advocates give advice based on their own judgment, not by simply passing along Gen AI output as that advice. The guidance prohibits delegating professional judgment, legal advice, or advocacy entirely to AI, including letting an autonomous AI system file documents or communicate with a tribunal without an advocate’s review.
Protecting Client Interests When Selecting AI Tools
Before adopting a tool, law firms are advised to run vendor due diligence covering the provider’s policy on using input data to train its model, data retention and storage location, access controls, and independent security certification such as ISO/IEC 27001, ISO/IEC 27701, or ISO/IEC 42001.
Maintaining Human Supervision and Review
The guidance sorts AI use into three risk tiers, low, medium, and high. Medium-risk tasks such as drafting a memo require direct oversight (“human in the loop”). High-risk tasks, including delegating advice entirely to AI or filing documents with a tribunal without review, are prohibited outright.
Being Transparent With Clients Where AI Use Has a Material Effect
The guidance does not require disclosure every time an advocate uses AI. A duty to disclose arises in three specific circumstances, when the client asks, when the engagement letter requires it, or when the advocate intends to input confidential client information into a Gen AI tool with the client’s consent.
Establishing Internal Governance for AI Use
Law firms are advised to adopt a written policy covering governance structure, an approved tool list, the level of oversight required, and an incident-reporting procedure, including any notification obligation under the Personal Data Protection Law where personal data is involved in an incident.
Where the Responsibility Lands
For law firms, the guidance shifts the frame of reference. AI use is no longer just an individual advocate’s habit, it becomes an institutional responsibility. An advocate who occasionally uses a chatbot to tidy up a letter carries a very different risk profile from a firm that has woven AI into its entire workflow, from client intake through billing.
Firms must also keep mentoring trainee advocates under Article 29(6) of the Advocates Law, even where AI can now handle part of their research and drafting work. The efficiency AI offers cannot be allowed to erase the learning process that builds a young advocate’s core skills.
PERADI’s guidance does not bar advocates from using artificial intelligence. What has changed is the clarity of the line between work a machine may assist with and work that must stay in human hands. Verification, confidentiality, supervision, and disclosure are not extra paperwork. They are an advocate’s long-standing duties, now applied to a new technological context.
