Google, OpenAI, and Anthropic are advancing the creation of an autonomous organization intended to define safety standards for cutting-edge artificial intelligence (AI) systems. This initiative comes amid growing discussion about how increasingly sophisticated technologies should be regulated.
As reported by The Information, this safety group is taking shape, and the three companies have provisionally named the entity the Standards Authority for Frontier AI (SAFA). The intention is for this new structure to operate independently, helping to fill gaps in governmental AI oversight.
The idea of establishing more comprehensive guidelines gained prominence this week after OpenAI CEO Sam Altman and Anthropic CEO Dario Amodei debated AI safety at the United Nations Security Council (UN). Altman argued that crucial decisions regarding AI must be guided by democratic institutions and governments. Amodei, on the other hand, advocated for global standards and international agreements to test new AI models.
Additionally, OpenAI separately requested that the United States take the lead in international efforts to define technical standards for advanced AI systems, including those with recursive self-improvement capabilities.
Divergences on AI regulation
The suggestion of SAFA arises during a period marked by growing disagreements over the best way to regulate AI. On one side, there are proponents of stricter safeguards to mitigate potential risks associated with increasingly capable systems. On the other side, some argue that current legislation and market mechanisms are sufficient to address such threats.
The U.S. government is also participating in this debate. President Donald Trump has expressed opposition to new regulatory restrictions on AI, arguing that both the US and China should maintain technological development at the current pace.
In this landscape, the body proposed by Google, OpenAI, and Anthropic represents an attempt by the leading AI developers to standardize safety, while governments assess what degree of supervision should be imposed on the most advanced AI systems. The news about the union of Google, OpenAI, and Anthropic to create the safety body was initially published in Olhar Digital.