DICT and CSC AI circular: a person stays accountable
The government's new AI principles keep a person accountable for AI-assisted decisions and promise a human review. The same design suits any office.
Published · 4 min read

On 9 June the Department of Information and Communications Technology (DICT) and the Civil Service Commission (CSC) issued Joint Memorandum Circular No. 003, the principles every government office must follow when it develops, deploys or uses artificial intelligence. It covers national agencies, constitutional bodies, local government units, state universities and colleges, and government corporations with original charters. Its most useful line for anyone building these systems sits under human-centricity: where a decision could endanger lives or cause damage that is hard to reverse, AI may gather insights and make recommendations, "but the final decision shall always be made by a human." It is the same rule behind the AI I have built, and it works well beyond government.
What the circular asks of an agency
The circular sets out ten principles, drawn from the OECD, UNESCO and ASEAN, and then gets specific. "AI systems shall not displace ultimate human responsibility and accountability," it says, and an agency that outsources the work stays accountable for its third-party developers. Agencies that deploy AI must disclose whether it was used in their deliverables and services, language models included. The circular gives AI users, which it defines as the officials and employees it covers, the right to contest an AI decision and ask for a human review, and entitles affected individuals to redress when a decision is wrong. The law firm Cruz Marcelo reads the review right as covering anyone affected.
Each agency must also form an internal committee, with its IT and legal units at the least and the people who build and run the AI, to advise the head of agency on how far AI may go. Every covered agency must issue its own AI policy and write the circular into its information systems strategic plan, its procurement, its terms of reference and its contracts. The DICT and the CSC are to issue implementing guidelines within 90 days of the circular taking effect. And as Baker McKenzie noted in July, it has implications for the private companies that build or supply AI for government.
The AI proposes, a person confirms
I build approval systems, where nothing is final until the person accountable for it has signed it off, and the circular asks the same of AI. On Storia, the wedding-planning platform I run, the assistant can change a couple's budget, wedding date, timeline and guest list, but every change is shown to the couple first and saved only when they confirm it. I wrote about that design in The AI proposes, a person confirms.
In a government office the same shape would look like this. An AI that reads an incoming letter proposes where it should go and drafts a summary, and the officer confirms or corrects both before anything moves. The record keeps the proposal and the decision side by side, with the officer's name on the decision. That is what accountability needs in practice: when someone asks for a human review, there is a person, a proposal and a decision to review, rather than an output nobody owns.
Five things to settle before switching AI on
- Name the person who confirms each kind of AI-assisted decision, and have the system record who it was, not only that someone clicked.
- Keep what the AI proposed beside what the person decided, so a review has both to look at.
- Say where AI was used, on the screen and in the letter that goes out. The circular leaves each agency to set its method of disclosure in its own policy.
- Give people a route to challenge a result and reach a person, with a named owner and a time limit.
- Put it in the contract: if a supplier builds or runs the AI, the terms should say what it may do with your data, what it must tell you about its limits, and that the decision stays yours. The circular itself holds vendors accountable for failing to disclose AI risks and limitations.
The circular is written for government, but its checks fit any office that lets AI near its records. If you want AI to take on the reading and drafting while your staff keep the decision, that is what Automate it is for, and it starts with a pilot on your own documents. The fit check is the first step.
Sources
- Official Gazette, Joint Memorandum Circular No. 003, Series of 2026: Principles for the Ethical and Responsible Development, Deployment, and Use of Artificial Intelligence (AI) in the Government, 9 June 2026
- Cruz Marcelo & Tenefrancia, DICT and CSC Issue Joint Memorandum Circular Establishing Principles for the Ethical and Responsible Development, Deployment, and Use of Artificial Intelligence in the Government, 13 July 2026
- Baker McKenzie, Philippines Launches Responsible AI Governance Framework, 20 July 2026
Enrique Lacambra III builds internal business systems from Cainta, Rizal.
Tell me what is broken and I will tell you if it is worth building. A 20-minute call, no charge, and I will say so if you do not need me.
