Differing Positions on AI Governance and Guardrails Among Key Institutions

Recently, Anthropic, OpenAI, the Vatican, and the U.S. Congress have agreed that guardrails are needed for AI governance, but they show differences in the core values that should be protected.

The main perspectives of each institution are as follows:

• Anthropic: Focuses on preventing Catastrophic Risk, guarding against cyber and biological threats and the possibility of models becoming autonomous and uncontrollable. It emphasizes independent testing and incident reporting.
• The Vatican: Prioritizes Human Dignity above all, treating moral issues such as job loss, surveillance, autonomous weapons, and algorithmic discrimination as core challenges.
• OpenAI: Emphasizes National Capacity, valuing U.S. AI leadership along with infrastructure, energy, and talent acquisition. At the same time, it supports building a social safety net in response to labor market changes.
• U.S. Congress (draft Great American AI Act): Focuses on Implementation, aiming to require large developers to disclose risk management frameworks, publish transparency reports, and utilize independent verification bodies.

This trend suggests that AI is establishing itself as core infrastructure for the economy and national security, going beyond being a mere product.