en.Wedoany.com Reported - Naver has expanded the scope of AI safety management from the model itself to the AI services actually used by users, with the core being an extension from risk management of AI models to full lifecycle safety management covering service planning to operational deployment.

Naver unveiled its AI safety management system "ASF (AI Safety Framework) 2.0" at the "AI Safety Seoul Forum (SFASS)" held in Gangnam-gu, Seoul on the 8th.

ASF 2.0 is a revision of the ASF announced by Naver at the AI Seoul Summit in 2024. While the original version focused on the performance and risk management of AI models themselves, this upgrade extends management to the entire process from service provision to user usage, as well as the actual usage environment.

This revision reflects changes in Naver's AI service strategy. The company has restructured its management system in response to technological and regulatory environmental changes, including the "On-Service AI" strategy of applying AI across overall services such as AI search and shopping AI agents, the multi-model environment combining various AI models, and the implementation of the Artificial Intelligence Basic Act. The target of safety management is no longer a single AI model, but the entire AI service composed of multiple AI models combined.

ASF 2.0 categorizes potential risks in AI services through an "AI Risk Taxonomy" and evaluates the service's application domain, scope of use, and impact on users and society using an "AI Impact Assessment Matrix." Based on this, necessary safety measures are implemented, and risks are continuously managed through safety assessments and user feedback after deployment.

Naver also operates a company-level execution system called "CHEC 2.0 (Naver AI Ethics and Safety Consultation Execution System)," which applies the above standards to the actual service development process. This is a process that checks AI safety with consistent standards from service planning to operational deployment. The "AI Tab" unveiled in June this year also underwent safety checks through CHEC 2.0 from the design stage to deployment, and the company plans to gradually apply this system to upcoming or currently operating AI-based services.

In March this year, Naver established a dedicated AI safety organization, the "AI Safety Center," under the Chief Responsibility Officer (CRO). The company manages AI safety through a three-tier management system where execution, management, and supervision are handled by the service departments, the AI Safety Center, and the board of directors, respectively.

Song Dae-seop, head of Naver's AI Safety Policy, stated that as the technological, service, and policy environment surrounding AI evolves, it is becoming increasingly important not only to ensure the safety of individual models but also to safely design and operate services that combine multiple AI models. He said the company will continue to accumulate experience and know-how in AI safety and collaborate with academia, policy institutions, external experts, and other organizations to continuously share and develop this system.