Government to Introduce 'Watermark' on AI Creations to Enhance AI Trust and Safety

The government is promoting the institutionalization of 'watermark introduction' for results generated by artificial intelligence (AI). The aim is to minimize the side effects that may arise from AI-generated content such as deepfakes and fake news.

On the 25th, Minister Lee Jong-ho of the Ministry of Science and ICT announced this plan at the '4th AI High-Level Strategic Dialogue' held at LG Science Park in Gangseo-gu, Seoul.

Minister Lee said, "AI is a core technology in the era of digital intensification, and efforts to secure reliability and safety are more important than anything else," adding, "We will establish an AI ethics ecosystem foundation including support for private autonomous AI reliability verification and certification, and develop an appropriate regulatory system."

Accordingly, the government will promote a plan to secure AI ethics and reliability. The plan includes ▲support for private autonomous AI ethics and reliability assurance ▲establishment of world-leading AI ethics and reliability technological and institutional foundations ▲spread of responsible AI awareness throughout society, aiming to become a model country for AI ethics and reliability.

As a detailed task, the government proposed the 'institutionalization of watermark introduction' for AI-generated results. This is intended to prevent side effects from AI-generated content such as deepfakes and fake news in advance and to protect the copyright of original content. In the first quarter of next year, a guidebook for 'high-risk AI' that significantly affects the safety of citizens' lives, bodies, and fundamental rights will be prepared, and related systems will be established.

However, the method of displaying the watermark on content has not been decided. It is expected to be a method that allows machines to verify AI-generated content rather than a visible mark to the human eye.

The government also plans to prepare and expand sector-specific guidelines for generative AI-based services and start private autonomous reliability verification and certification from next month. In particular, some companies conducting high-risk AI development and demonstration projects will be selected to pilot certification in December.

Furthermore, to respond to potential risk factors such as bias and opacity inherent in AI itself and AI malfunctions, new technology development to secure reliability will be promoted from next year. A budget of 22 billion KRW will be invested over three years until 2027.

Meanwhile, at the high-level meeting, representatives from leading domestic AI industry companies such as Naver, Kakao, SK Telecom, KT, and LG AI Research Institute attended and held discussions on the necessity of AI reliability verification and certification.

Kim Yoo-cheol, head of LG AI Research Institute, gave a presentation titled 'Corporate Efforts to Implement AI Ethics Principles.' Following this, participants discussed three topics: ▲directions for promoting private autonomous verification and certification to balance AI industry development and reliability assurance ▲voluntary efforts to improve the reliability of large-scale and generative AI ▲suggestions for spreading AI ethical awareness and leading global norms.

The participants agreed on the necessity of introducing watermarks for AI-generated results and consented to gradually apply them to their services considering technical completeness.

This content was produced with the assistance of AI translation services.