Sam Altman: Controlling the pace of AI development doesn't mean stopping, but capabilities must not advance faster than safety alignment and monitoring

Odaily News: OpenAI CEO Sam Altman posted on X, welcoming the establishment of a federal-level frontier AI safety regulatory framework in the United States to set consistent safety requirements for the industry, but noted that AI companies don't need to wait for antitrust exemptions or related legislation to take proactive safety measures.

Altman stated that OpenAI currently develops clear safety justifications in advance before starting frontier reinforcement learning training that is expected to significantly enhance model capabilities, and hopes the industry will form common standards on model失控 risk, monitoring, and safety. He emphasized that so-called "pacing" of AI development does not mean stopping progress, but rather accepting the speed and cost trade-offs brought by safety measures: "No matter how much competitive pressure the United States faces, AI capabilities should not be allowed to develop faster than alignment and monitoring capabilities."

Previously, Anthropic, OpenAI, and Google had been engaged in working group consultations since July on establishing their own AI safety standards organizations. Additionally, U.S. Democratic lawmakers will hold a caucus meeting on Tuesday morning to specifically discuss legislative actions related to artificial intelligence, stating that addressing the challenges posed by AI is a "high priority" requiring prompt measures.