Abstract:
OpenAI urges the United States to take the lead in developing standards for cutting-edge artificial intelligence (AI) with other countries in response to growing concerns about the potential harm of AI. OpenAI said in a blog post on Monday that in order to develop artificial intelligence safely, it is necessary to develop international technical standards, especially to help define what types of incidents involving AI models should be notified, to clarify how countries can cooperate in AI development, and to improve access to the computing power needed to develop such technologies.
“To make continued progress in AI safe and beneficial, it is necessary to advance aligned research and develop common standards to provide guidance for AI development in laboratories and countries,” OpenAI said.
OpenAI said its CEO Sam Altman plans to speak to the United Nations Security Council on Wednesday on AI and international security issues.
U.S. Treasury Secretary Bessent said last weekend that the United States proposed to establish a reporting mechanism to cover AI incidents with national security implications.
OpenAI emphasized in a blog post that countries should work together to develop complementary national and international standards for developing the most capable AI models. OpenAI suggests that this work can be promoted through existing AI safety agencies in multiple countries, including the Center for AI Standards and Innovation (CAISI) under the U.S. Department of Commerce, which assists in testing AI models and developing relevant guidelines.
OpenAI also called for the development of standards to deal with the increasing ability of AI to work autonomously, noting that "fully autonomous" recursive self-improvement - where the AI is ultimately able to modify its own code or improve its capabilities without any human intervention - should not be pursued "unless and until safety can be ensured."
Comments