OpenAI and Anthropic Reach Agreement: U.S. Government to Conduct Early Safety Assessments on New AI Models

OpenAI and Anthropic Sign AI Safety Cooperation Agreement

Recently, renowned AI companies OpenAI and Anthropic announced a significant cooperation agreement with the U.S. AI Safety Institute. Under this memorandum of understanding, both companies pledge to provide the U.S. government with early access to new AI models for security evaluations before their public release. This collaboration represents a crucial advancement in the safe development of AI technology.

According to the agreement, both OpenAI and Anthropic will grant the U.S. government access to their AI models both prior to and following their public launches. This initiative aims to help identify and mitigate potential safety risks. The U.S. government has indicated that this partnership will enhance the understanding and regulation of AI technologies, ensuring their safe and controllable development.

OpenAI’s Chief Strategy Officer, Jason Kwon, expressed strong support for the initiative, highlighting the importance of working with the U.S. AI Safety Institute to establish best practices and safety standards for AI models. Kwon believes this collaboration will bolster the United States' leadership in the responsible development of AI.

Jack Clark, co-founder and policy lead at Anthropic, also expressed optimism about the partnership, emphasizing the significance of ensuring the safety and reliability of AI technologies in making a positive impact. Through this collaboration, both organizations aim to better identify and reduce the risks associated with AI, promoting responsible AI development.

This announcement comes at a time of heightened global discussions about AI regulatory policies. As legislative bodies at various levels explore how to balance innovation with necessary safety regulations, the OpenAI-Anthropic partnership sets a benchmark for the industry. Recently, California passed the "Advanced AI Model Safety Innovation Act" (SB 1047), which mandates specific safety measures for AI companies before training advanced models. Although the bill awaits the governor's signature, it has already sparked widespread attention and debate.

The White House is actively seeking voluntary commitments from leading AI companies concerning safety measures. Several top firms have expressed non-binding pledges, investing in cybersecurity and fairness research, and committing to watermarking AI-generated content for improved safety.

Elizabeth Kelly, Director of the U.S. AI Safety Institute, remarked that while these new agreements are just the beginning, they represent an essential milestone in managing the responsible future development of AI. She is committed to promoting the safe advancement of AI technologies to ensure their benefits to society.

The collaboration between OpenAI, Anthropic, and the U.S. government injects new momentum into the safe development of AI technology. As AI continues to advance and its applications diversify, ensuring the safety and trustworthiness of AI will become a focal point of global concern. By strengthening cooperation and regulation, AI technologies are expected to play an increasingly positive role in driving societal progress and development.