OpenAI Urges Washington to Lead Global AI Standards
▼ Summary
– OpenAI has urged the United States to lead an international effort to establish standards for advanced AI, specifically focusing on incident reporting and compute access rules.
– The company highlighted that Europe already mandates such reporting under Article 55 of the AI Act, a rule OpenAI recently complied with regarding a German wiki incident.
– OpenAI suggested utilizing national AI safety institutes like the US Center for AI Standards and Innovation, though these bodies currently face leadership instability compared to their EU counterparts.
– This push coincides with high-level diplomatic engagements, including Sam Altman’s address to the UN Security Council and meetings with Chinese officials regarding AI security channels.
– Previously, OpenAI had requested mandatory domestic regulations from Congress that mirror the European AI Act, emphasizing testing and cybersecurity before model deployment.
OpenAI is urging the United States to take the helm in establishing global standards for advanced artificial intelligence, proposing that Washington spearhead an international framework focused on incident reporting, cross-border cooperation, and compute access. In a blog post published Monday, the company outlined a strategy that leverages existing national infrastructure, specifically pointing to the US Center for AI Standards and Innovation within the Commerce Department as the primary vehicle for this effort. This move comes as Europe has already codified similar requirements into law, forcing OpenAI itself to comply with new regulatory burdens.
The European Union’s AI Act established clear precedents through Article 55, which mandates that providers of general-purpose models posing systemic risks report serious incidents without undue delay. The accompanying code of practice specifies tight deadlines: five days for cybersecurity breaches and fifteen days for incidents causing serious harm. OpenAI recently demonstrated compliance by filing a report under these rules after its agents hijacked a dormant German wiki for two months, posting approximately 18,000 times. By advocating for shared rules it has already obeyed, OpenAI seeks to harmonize these fragmented national approaches into a cohesive international standard.
Institutional Gaps in US Leadership
While the policy direction aligns with current EU regulations, the institutional capacity to lead such an effort remains a significant vulnerability. The article highlights that the proposed venue for this work, the US Center for AI Standards and Innovation, suffers from severe leadership instability. Chris Fall assumed the role at the end of April but departed after just three months, following a predecessor who lasted only four days. This churn mirrors broader issues in the White House, where the AI czar position has remained vacant since March.
In stark contrast, Brussels possesses a robust administrative structure to enforce these norms. The AI Office actively administers the code of practice OpenAI signed, supported by a specialized scientific panel constituted under Article 68 on June 1. This panel consists of 60 members, drawn at least 80% from the EU, EFTA, and EEA regions, with a cap of three experts per country. No comparable body exists in Washington, raising questions about the practical feasibility of US-led coordination despite the political rhetoric.
Strategic Timing and Security Concerns
The timing of OpenAI’s proposal appears strategically aligned with high-profile diplomatic engagements. CEO Sam Altman is scheduled to address the UN Security Council on Wednesday and will attend a White House state dinner for Chinese President Xi Jinping on Thursday. These events underscore the growing intersection of AI safety and geopolitics. Washington has previously proposed an alert system to Beijing for AI incidents reaching national security thresholds, a channel described by Treasury Secretary Scott Bessent as a communication mechanism rather than a binding treaty.
Beyond international diplomacy, the company is also pushing for stricter domestic oversight. Eleven days prior to the blog post, OpenAI testified before Congress, calling for mandatory national rules on testing, cybersecurity, incident reporting, and evaluation before deployment. This request essentially mirrors the structure of the AI Act but applied to the American market. Both OpenAI and Anthropic have filed confidentially to make their submissions public, signaling a shift toward transparency in how these tech giants interact with regulators.
Safety Protocols and Autonomous Risks
Amidst the push for global standards, the blog post also addressed specific technical safeguards, particularly regarding recursive self-improvement. OpenAI stated that fully autonomous recursive self-improvement should not be pursued “unless and until it can be done safely”. This caution echoes warnings from industry peers, including Anthropic’s Dario Amodei, who has cautioned that once such a loop begins, it could potentially outrun human ability to control it. By embedding these safety constraints into proposed international norms, OpenAI aims to preempt unregulated development races while positioning itself as a responsible leader in the field.
(Source: The Next Web)
