[123RF]
[123RF]

Organizations deploying AI services related to energy or drinking water will be required to verify that they have management systems in place allowing humans to immediately intervene or shut down those services.

According to an AI ethics self-checklist released Tuesday by the Ministry of Climate, Environment and Energy, any "high-impact AI" service introduced for energy supply or drinking water production must be accompanied by a management framework enabling "immediate human intervention, suspension and review."

Before deployment, operators must assess the potential impact of the AI system in advance and confirm that response mechanisms — including prevention, mitigation and recovery measures — are in place.

High-impact AI refers to systems that could significantly affect human life, physical safety or fundamental rights, or that carry the risk of causing serious harm.

The ministry's checklist adapts the "Public Sector AI Ethics Principles" issued by the Ministry of Interior and Safety in November last year to the ministry's specific mandate. The public-sector principles set out different requirements depending on the level of AI autonomy involved.

The requirement for a human intervention and override framework applies specifically to Level 3 deployments — cases in which AI is given autonomous decision-making authority.

For all AI service deployments, the ministry will require verification that the system contributes to advancing "environmental values such as carbon neutrality, climate crisis response and energy transition" as well as "effectiveness in climate disaster response, environmental management and energy supply and demand."

Checklists will also examine whether AI services have taken into account not only the current generation but future generations as well.

When sensitive information is used, operators must confirm that relevant laws and procedures have been followed. The checklist identifies "environmental and public health information," "personal energy consumption patterns" and "location data" as examples of sensitive information.

For Level 3 AI services, additional checks will be required to confirm that the system contributes to "enhancing sustainability in the climate, energy and environment sectors" and "protecting future generations," and that it accounts for "environmental impacts such as energy use and greenhouse gas emissions" as well as "conflicts with stakeholders."

Consideration must also be given to whether the environmental benefits generated by AI deployment are distributed equitably to climate-, energy- and environment-vulnerable groups and regions.

The ministry's AI ethics self-checklist is scheduled to take effect in October.


thlee@heraldcorp.com