This is a summary of Day 3 of EngageMedia’s Training on Strategic Engagement with Government regarding AI Governance and AI Accountability, held in Bali, Indonesia, from May 13 to 15, 2025. Read the recaps for Day 1 and Day 2.
Day 3 of EngageMedia’s workshop shifted course into AI governance and policy in the global and local contexts (Indonesia and Malaysia) and how the human rights approach should be the baseline of policymaking.
Global overview on AI governance
Shabnam Mojtahedi from ICNL led a session breaking down the international bodies responsible for the governance and international regulations over AI and emerging technologies. Among these bodies are the United Nations, the European Union and Council of Europe, and international standards-setting bodies such as ISO, IEC, and IEE.
Mojtahedi delves also into differences of principles in governing states and corporations; where states are obligatorily bound to uphold the norms under the agreements they have signed on – such as the Universal Declaration of Human Rights – corporations, especially those existing in a large-scale, multinational form, cannot be bound explicitly by international agreements. Thus, the global approach towards regulating AI at the corporate level falls on guidelines – for example, the UN Guiding Principles on Business and Human Rights – which are not legally binding, but are influential and impactful to some extent on the internal policies of companies. Still, there needs to be more effort to ensure that companies and states actually follow through on the principles they say they’re committing towards.
Human rights impact assessment
One way to ensure and measure commitment to the principles is via a Human Rights Impact Assessment (HRIA) Audit. The HRIA methodology varies from one entity to the other, but it essentially assesses the following aspects of a technology that we want to implement: Rights Respecting; Trustworthiness; Reliability; Robustness; and Accuracy.
Rachel Judhistari from Wikimedia Foundation outlined the example of applying HRIA in Wikimedia. The organization has been using AI as an assisting tool for their human contributors, and the community governed AI through decentralized multi-stakeholder models which are field-tested to ensure information integrity and the ethical use of AI.
In Wikimedia, the AI tools are used for the following purposes, among others: supporting Wikipedia’s moderators and patrollers for protecting against vandalism; preserving indigenous language and practices featured in Wikipedia; and monitoring the validity and relevance of the content or proposed changes. The HRIA Audit in Wikimedia is therefore based on its Human Rights Policy formed in 2021, which emphasizes independence throughout all its phases and an all-stage evaluation, which means that the systems are evaluated even when they’re not flagged as high-risk.
Pic. 1: One of the group works on simulating the human rights impact assessment exercise.
The classroom then broke out into group work — facilitated by Mojtahedi and Judhistari — to design an audit framework based on the scenario provided by trainers. One interesting commonality that came up in the group is the importance of the inclusion of linguistic experts to evaluate the systems and contents based on the unique, cultural nuances it wishes to address and represent. This is especially important because most AI systems are made with English as their primary languages, which may not account for the linguistic variety of other languages they serve.
AI governance in Malaysia context
Dr. Jun-E Tan from the Khazanah Research Institute then continued to share current updates on AI governance in Malaysia. She first explained the key players in the sector. The three key ministries are the Ministry of Science, Technology and Innovation (MOSTI), Ministry of Digital (KD), and the Ministry of Communications (K-KOMM). These ministries collaborate with bodies tasked with executing and innovating implementations of AI, such as SIRIM, which is involved in AI in R&D and knowledge development in Malaysia, and the National Cyber Security Agency (NACSA), which is responsible for the security measures around adapted tech and AI.
She also elaborated on the rapid speed of AI development picking up in 2024, with big goals ahead in 2025. The newly-established NAIO office, for example, has six deliverables on Action Plans, Regulatory Framework, Adaptation and Adoption Guideline, AI Ethics, Impact Studies, and Data Governance, all due at the beginning of 2026.
Pic. 2: Malaysia’s Timeline for AI Governance and AI Adaptation, from 2021 – 2024.
She also mentioned notable developments in regulatory and sectoral guidance, public awareness and human capital, and participation in international governance. One example is the development of the AI Governance and Ethics (AIGE) Guidelines, launched by MOSTI in September 2024, which provide guidance for end users, policymakers, as well as developers and technology providers. The guideline is used by other agencies to develop sector-specific codes of practices.
Indonesia National AI roadmap 2025
Meanwhile, in Indonesia, Mr. Ridwan from the Ministry of Communications and Digital Indonesia (KOMDIGI) shared the current developments for AI governance in the country in accordance with the National AI Roadmap 2025 – 2045. He laid out the regulatory foundation of present AI integration in governance and business, namely the Circular Letter no. 9/2023 on AI Ethics, and the plans to expand AI regulations through Presidential Regulation (Peraturan Presiden). He clarified that the government chose Presidential Regulation over Legislative Enactment (Undang-Undang) for the short time it needs to be formed and enforced, noting that speed is an essential factor in regulating emerging technology. He also highlighted existing legislative enactments that will support the governance of AI, such as the Data Protection Act (2022) and the Amendment to the Electronic Information and Transactions Act (2024).
Pic. 3: Presentation from the Ministry of Communications and Digital Indonesia on the Indonesia National AI Roadmap 2025.
Presently, the government is focusing on supporting AI integration in several sectors, with priorities in the energy and mineral sector, fisheries, agriculture and farming. The Ministry is also focusing on the deployment of analytic and predictive AI, Digital Twin, Internet of Things and Enterprise Asset Management, Distributed Ledger, and Process Automation in relevant fields to increase productivity and efficiency in governance and government-owned corporations. The government is also particularly interested in the integration of LLMs to assist in its public-facing services and communications.
The government is also expected to deliver new regulatory developments on AI by 2025, and is already forming working groups for AI Ethics, Infrastructure & Data, Talent Development, R&D, and Funding & Investments.
Pic. 4: Indonesia’s Timeline for AI Regulation Development in 2025
Ridwan acknowledges that presently, many of these Working Groups are dominated by representatives of business entities, particularly Big Tech, and are focusing more on the integration of AI in the context of business. He invites concerned civil society actors to provide their feedback and insights to the respective working groups by July latest.
UN mechanisms and other initiatives
Nento Meyda from the United Nations Educational, Scientific and Cultural Organization (UNESCO) elaborated on the steps taken at the global level to develop regulatory and ethical guidelines for AI adoption in various member states. Under its mandate to promote international cooperation in education, science, culture, and communication to foster peace and equality worldwide, the organization first adopted the Recommendations on the Ethics of AI by acclamation of 193 Member states on November 23, 2021. The Recommendation acknowledges the responsibility to develop the following;
| A dynamic assessment method that evaluates the readiness of countries in developing, implementing and using AI, which is continuously evolving. | Which leads to | Readiness Assessment Methodology (RAM) Macro Diagnostic Tool Helps countries to understand where they are on the scale of preparedness in ethical and responsible AI adoptions for public interest Highlights the institutional and regulatory changes needed domesticallyHelps UNESCO tailor capacity-building efforts to country-specific needs |
| A methodology to detect and address the challenges and shortcomings of AI against the agreed-upon Ethical principles | Which leads to | Ethical Impact Assessment (EIA) Micro-level procurement tool Helps the government official responsible for AI procurement to determine which system to purchase Highlights alignment of AI systems with the ethical standards set out in the UNESCO Recommendation |
Further, Karlina Octaviany from GIZ outlined the various efforts made by their Project FAIR FORWARD to integrate AI within various public-interest initiatives. FAIR FORWARD aims to democratize AI worldwide by fostering local innovation by Open Sourcing AI, which includes:
- Improving access to training data and open AI technologies for local innovation
- Strengthening local technical know-how on AI in Africa and Asia
- Developing policy frameworks for ethical AI &
- Fostering dialogue on AI and Data Protection
Notable innovations in the field of LLM integration include their collaboration with Prosa.ai, which developed text datasets, language models and benchmarks for three regional languages in Indonesia — Minangkabau, Balinese, and Buginese; and explored the issues of misinformation and disinformation related to the climate crisis in Indonesia.
The objectives of the Day 3 session were to explore the utilization of a human rights framework in the AI policy-making process and to examine how civil society members can effectively and strategically position themselves in engaging with policymakers.
