Responsible AI in Philippine government agencies

The Philippine government now has a common rulebook for its own use of AI. Joint Memorandum Circular No. 003, s. 2026, titled Principles for the Ethical and Responsible Development, Deployment, and Use of Artificial Intelligence (AI) in the Government, was issued by the Department of Information and Communications Technology (DICT) and the Civil Service Commission (CSC) on 9 June 2026. It takes effect on compliance with publication requirements, and the DICT and CSC are to issue implementing guidelines within 90 days of that. This guide sets out what the circular asks and how an agency can set up the work.

Who it covers

The circular applies to government officials and employees in constitutional bodies, national departments, bureaus and agencies, government-owned or controlled corporations with original charters, local government units and state universities and colleges, whatever their appointment status. Agencies must also make sure that people engaged under job orders or service contracts whose work involves AI comply.

The definition of AI is broad, covering machine learning, generative AI, autonomous systems and similar technologies that produce recommendations or decisions. Outsourcing does not move responsibility. An agency that procures an AI system or service remains accountable for the decisions it supports.

The ten principles

The circular sets ten foundational principles, drawn from OECD, UNESCO and ASEAN frameworks.

1. Proportionality and do no harm
2. Human-centricity
3. Fairness and non-discrimination
4. Transparency and explainability
5. Data governance, privacy and protection
6. Responsibility and accountability
7. Multistakeholder and adaptive governance
8. Cyber safety and security
9. Robustness and reliability
10. Sustainability

The principles turn into concrete expectations. Risk assessments and safeguards where risks exist. Human oversight of high-risk decisions. Bias prevention, including bias audits. Clear disclosure of an AI system’s purpose, inputs, risks and limitations. And privacy impact assessments, in line with the Data Privacy Act.

Duties by role

The circular assigns duties by role. Developers are responsible for responsible design, testing and documentation, performance monitoring and data integrity. Deployers must disclose AI use in government deliverables and services. Users remain accountable for how they use AI. Vendors must provide reliable, secure and ethical solutions and may be held accountable for flawed products or undisclosed material risks. And persons affected by AI-assisted decisions may contest them and request human review.

Set up the agency AI committee

Each covered agency that uses AI must form an internal committee. At a minimum it includes IT and legal representatives, plus representatives of the AI developer and deployer. It advises the head of agency on what AI development, deployment and use is permissible, and the head of agency decides.

A committee can only advise on what it can see. Its first need is a list of the AI the agency already uses, including generative AI tools staff have adopted and AI features inside systems the agency runs. Its second is a standard way for new AI to come to it before it is used. The AI system register guide and the AI use case intake guide cover both.

Build it into plans, policy and procurement

Agencies must adopt or update internal AI governance policies, and build the circular’s requirements into their Information Systems Strategic Plans, procurement, technical specifications, terms of reference, service agreements and contracts. Government data used in AI must be handled under the Data Privacy Act, cybersecurity issuances and records management policies.

For most agencies the policy work is an AI acceptable use policy for staff and a governance policy that sets out the committee’s role, how AI is approved and how affected people can seek review. The acceptable use policy guide is a starting point. The procurement work is a set of contract terms that make vendors disclose what their AI does, its risks and limitations, and material changes. The vendor AI risk guide covers what to ask for.

Make human review real

The right to contest an AI-assisted decision and ask for human review only means something if the agency knows which decisions AI touches and who reviews them. Record, for each system, whether it affects people, what it decides or recommends and where a person reviews its output. The same record supports the privacy side, because the National Privacy Commission’s Advisory No. 2024-04 expects meaningful human intervention for AI decisions with significant effects, and the Data Privacy Act gives people a right to information on automated decisions.

Where Aicura fits

Aicura’s AI Register holds every system and agent with its owners, vendor, outputs, whether it affects individuals and its level of human oversight, and discovery surfaces AI in use that is not yet on it. Risk assessments run when a system is registered and again on material change, and governance policies are drafted and kept current. Aicura does not assess against Philippine government issuances. Its register, assessments and policies support the work the circular describes, and the decisions stay with the agency’s committee and head.

A note on this page

This is Aicura’s reading of JMC No. 003, s. 2026, based on public reporting of the circular. It is not legal advice. Read the circular as published by the DICT and the Civil Service Commission, and follow its implementing guidelines once issued.