OpenAI has urged the United States to spearhead a global initiative to establish standards for advanced artificial intelligence, focusing first on the development of regulations that specify which AI-related incidents require reporting. The organization proposed that this effort be managed through national AI safety agencies, notably the US Center for AI Standards and Innovation. This comes in the wake of Europe’s implementation of similar legislation within its AI Act.
In a blog post highlighted by Bloomberg, OpenAI called for a unified framework that outlines incident reporting protocols, international cooperation, and data access regarding AI technologies. Europe has already set a precedent with Article 55 of the AI Act, which mandates that providers of general-purpose AI models report significant incidents promptly. The guidelines stipulate a reporting deadline of five days for cybersecurity breaches and fifteen days for serious harms. OpenAI recently complied with this requirement, informing authorities about AI agents that engaged with a dormant German Wikipedia site for two months, resulting in approximately 18,000 posts.
OpenAI noted the necessity of a rule for incident reporting it has already been following. The organization’s blog post suggested that national AI safety institutes should oversee this initiative, specifically naming the US Center for AI Standards and Innovation, which operates under the Department of Commerce. Additionally, OpenAI cautioned against pursuing fully autonomous recursive self-improvement until it can be guaranteed to be safe. Dario Amodei from Anthropic has previously warned that such self-enhancement could quickly surpass human control.
However, the US Center currently faces challenges, lacking a permanent director since Chris Fall’s brief tenure ended in July. His predecessor lasted just four days, and the position vacated by the White House AI czar in March remains unfilled. In contrast, Brussels has made significant strides; the AI Office oversees a code of practice that OpenAI has signed, and a 60-member scientific panel was appointed under Article 68 on June 1. This panel comprises at least 80% of members from the EU, EFTA, and EEA, with a cap of three experts per country. Washington has yet to establish a similar structure.
The timing of OpenAI’s appeal aligns closely with significant events, as CEO Sam Altman is set to address the UN Security Council and attend a White House state dinner for Chinese President Xi Jinping. The US government has already suggested a system to alert Beijing regarding AI incidents that may affect national security, with Treasury Secretary Scott Bessent describing it as a communication channel rather than a formal treaty.
OpenAI's earlier request focused on domestic regulations, advocating for mandatory national rules about testing, cybersecurity, incident reporting, and evaluation prior to AI deployment. This proposal closely mirrors the AI Act but tailored to an American context. Both OpenAI and Anthropic are also reportedly preparing for confidential public offerings.
