TLDR: Major artificial intelligence companies, including Google, Meta, Microsoft, and OpenAI, gathered in Seoul on September 15, 2025, for an ‘Open-Source Day’ event. Hosted by the Personal Information Protection Commission, the discussions focused on the secure and reliable implementation of open-source AI, tackling critical issues like security, privacy, and the sharing of safety tools. The event highlighted a growing industry commitment to responsible AI development, with companies showcasing their platforms and strategies for mitigating risks.
Seoul, South Korea – On September 15, 2025, a significant assembly of global artificial intelligence leaders, including Google, Meta, Microsoft, and OpenAI, convened in Seoul to deliberate on the safe and responsible deployment of open-source AI. The ‘Open-Source Day’ event, organized by the Personal Information Protection Commission as a precursor to the Global Privacy Assembly (GPA), drew over 120 participants. Attendees included representatives from prominent open-source model and solution providers, domestic AI corporations, researchers, and international supervisory bodies.
The discussions underscored the increasing importance of addressing security and privacy concerns inherent in open-source AI. Participants shared insights and tools aimed at ensuring the reliable and safe implementation of these rapidly evolving technologies.
A pre-event survey conducted by the Personal Information Protection Commission revealed compelling statistics regarding the adoption and safety considerations of open-source AI. Out of 70 responding developers, researchers, and corporate officials, approximately 62% reported having experience in adopting and utilizing open-source solutions. Furthermore, a substantial 77% indicated that they actively considered safety aspects when fine-tuning models with open-source components.
Key corporations presented their approaches and contributions to the open-source AI ecosystem:
Google showcased its Vertex AI platform, designed for the cost-efficient operation of open-source models. The company also detailed tools for enhancing reliability and safety, such as LLM quality evaluation features, prompt optimization capabilities, and general safety enhancement tools.
Aim Intelligence provided real-world case studies, highlighting the diverse safety and information security challenges faced by corporations in operating AI services and internal AI models. Notably, Aim Intelligence successfully adapted Meta’s open-source AI filtering model, Llama Guard, for Korean conditions, earning them the ‘Llama Impact Innovation Awards.’
Microsoft presented a customer success story involving the development of agent AI using its Azure AI Foundry platform. The company emphasized the potential of leveraging open-source models and tools in the creation of agent AI, a paradigm gaining significant industry attention.
Naver introduced a suite of tools for safe open-source utilization, including public datasets, benchmarks, and an AI safety framework. They also shared their open-source model, HyperCLOVA X, and outlined past efforts to foster the growth of the domestic open-source ecosystem.
Also Read:
- Samsung Kicks Off Ninth Annual AI Forum, Unveiling Future of Artificial Intelligence and Semiconductor Innovation
- Global AI Leaders Convene in Shanghai to Chart Future of Innovation and Investment
OpenAI discussed the economic and social value of its recently released open-source model. The company also brought to light the challenges associated with the widespread adoption of open source, particularly concerns regarding safety and accountability, stressing the need for global-level discussions to address these issues.
