Ministry of Interior and Safety presents plan at Cabinet meeting Tuesday to boost AI development by civil servants, with performance rewards and liability protection for rule-following experiments

[123RF]
[123RF]

The government plans to spread a culture across ministries in which civil servants use AI to solve problems on the ground directly, with verified results taken up and shared by their organizations.

The Ministry of Interior and Safety said it presented a plan titled "Measures to Promote and Spread Direct AI Development by Civil Servants" at a Cabinet meeting Tuesday.

The plan is a follow-up to the "Strategy for Realizing the World's Best AI Democratic Government" announced Aug. 25, specifically fleshing out the "competent government for the people" pillar, which aims to transform how the public sector works.

The plan links every stage of direct AI development — from talent cultivation and development environments to security and quality verification, workplace application, pan-government rollout, and performance rewards and liability protection — into a single policy framework.

First, the government will expand the ranks of "AI champions" — civil servants who use AI to solve administrative problems hands-on — to 20,000, or about 2 percent of all staff across administrative and public institutions, by 2030.

The "AI Government Lab" will be expanded from its current internet-network focus to cover internal government networks, and will provide resources that are difficult for individuals to obtain on their own, including AI coding tools, development environments and GPUs. The number of simultaneous users is set to grow from the current level of about 200 to 6,000 by 2027.

Public data and administrative services will be made available through standardized interfaces such as APIs and MCP, so civil servants can plug them directly into their development work without having to rebuild data and functions from scratch.

Development outputs — including task documentation, source code, prompts, data specifications and test results — will be registered and systematically managed on a public GitLab platform.

Support will be provided in stages, from experimentation and verification through pilot operation and pan-government rollout.

Direct-development outputs will be managed across three stages: experimentation and verification, operational management, and formal IT projects. In the experimentation and verification stage, civil servants will define a workplace problem and build a prototype, with the ability to get started through minimal self-review — no prior deliberation or separate project-plan approval required.

In the operational management stage, the responsible department or agency will confirm necessity, scope of use, accuracy, security and personal data protection, and take responsibility for operations. The Ministry of Interior and Safety plans to centrally manage tasks and functions that multiple agencies can share, and to convert larger projects or services aimed at the public into formal IT projects to be run responsibly by the relevant agency.

Strong performance will be rewarded, and experiments that follow established procedures will be protected.

A reward system linking personnel evaluations, commendations and learning support will be put in place to ensure civil servants' AI development work does not remain a one-off effort. AI champions holding Black certification — the highest skill tier — will receive bonus points in the performance review for the Grade 5 early-promotion program's advanced science and technology talent category. The government also plans to expand incentives for top AI performers, including special performance bonuses, training and study opportunities, and additional learning subsidies.

Beyond individual rewards, the plan also aims to build organizational momentum for AI transformation, or AX.

An AX category will be added to innovation case competitions such as the "Government Innovation Grand Championship" and "First and Best" certification programs, and AX will carry greater weight in next year's government innovation ministry evaluations — all designed to encourage each agency to take the lead in building an AI-driven work culture.

Experiments that comply with established procedures and standards will be linked to the "proactive administration liability exemption" regime. Civil servants will be shielded from personal liability when there is no willful misconduct or gross negligence, giving institutional protection to new attempts.

Leadership training and hands-on workshops on AI will also be offered to department and division heads to help managers channel frontline experimentation into organizational results. In November, the government will hold the "2026 Public AI Grand Transformation Challenge" to identify and share best practices.

In the second half of this year, the government will work to establish guidelines for civil servants' direct AI development and output management, along with safety-check standards covering software vulnerabilities and personal data. In 2027, it plans to build an internal-network AI Government Lab and a security inspection system to support the safe use of direct-development outputs in the workplace.

Minister of Interior and Safety Yun Ho-jung said the competitiveness of an AI democratic government comes from individual civil servants — those who know the public and their work best — solving problems on the ground directly with AI. "We will ensure that civil servants' ideas can be tested quickly and freely, that organizations take responsibility for security, quality and operations when applying them to real work and public services, and that verified results are shared so every agency can use them," he said.


thlee@heraldcorp.com