OpenAI has urged the United States to spearhead a worldwide initiative establishing standards for advanced AI systems, according to a blog post released Monday and reported by Bloomberg. The company's primary request centers on harmonized protocols for reporting incidents tied to AI models, alongside frameworks for international cooperation and compute access.

Europe has already legislated the first component. The EU's AI Act mandates that providers of general-purpose models carrying systemic risk disclose serious incidents promptly—specifically within five days for cybersecurity breaches and fifteen days for cases of serious harm, as outlined in Article 55. OpenAI itself filed such a report this month regarding autonomous agents that remained active on a dormant German wiki for approximately two months, generating roughly 18,000 posts.

OpenAI proposes channeling this work through existing national AI safety institutes, specifically naming the US Center for AI Standards and Innovation, which operates under the Commerce Department. The company also advocates against pursuing fully autonomous recursive self-improvement "unless and until it can be done safely", echoing concerns previously raised by Anthropic's Dario Amodei about the risks of uncontrollable feedback loops.

The institutional challenge

The proposed American venue faces significant structural obstacles. The Center for AI Standards and Innovation currently lacks a permanent director. Chris Fall assumed the position at the end of April but departed after three months, while his predecessor lasted only four days. The White House's AI czar position has remained vacant since March.

Brussels confronts the inverse situation: its institutional infrastructure already exists and functions. The EU's AI Office oversees the code of practice that OpenAI has already signed. Additionally, a scientific panel comprising 60 members was established on 1 June under Article 68, with at least 80 percent drawn from EU, EFTA, and EEA member states and a three-expert cap per country. Washington has constructed no comparable apparatus.

Strategic timing

The timing of OpenAI's initiative warrants scrutiny. Sam Altman is scheduled to address the UN Security Council on Wednesday and will attend a White House state dinner for Xi Jinping on Thursday. Washington has already proposed an alert mechanism with Beijing for AI incidents affecting national security, which Treasury Secretary Scott Bessent characterized as a communication channel rather than a formal treaty.

The domestic precedent

Eleven days prior, OpenAI presented Congress with a separate request for mandatory domestic regulations covering testing, cybersecurity, incident reporting, and pre-deployment evaluation. This proposal essentially mirrors the EU's AI Act structure under an American framework. Both OpenAI and Anthropic have submitted confidential filings on this matter with plans to make them public.

Source: The Next Web