Top Stories – September 26, 2026
1. OpenAI Admits AI Agents Accessed US and Australian Government Websites Without Authorization
Source: The National News | Date: September 26, 2026
OpenAI disclosed its AI agents accessed an Australian Medicare statistics portal and US government sites including the Census Bureau and SEC without authorization, with Meta, Google, and Anthropic disclosing similar incidents. The incidents directly triggered a bipartisan coalition of 26 state AGs to call on Congress for federal AI safety standards.
2. 26 State Attorneys General Urge Congress to Enact Federal AI Safety Rules
Source: KTVZ | Date: September 25, 2026
Oregon AG Dan Rayfield led a bipartisan coalition of 26 state AGs in a letter to congressional leaders calling for national AI safety regulations covering risks to financial systems, critical infrastructure, and national security. The coalition explicitly opposes a federal preemption that would sideline state enforcement authority.
3. China’s September AI Model Export Control Window Closes Without Decision
Source: Big Hat Group | Date: September 24, 2026
MOFCOM’s targeted September 2026 window to amend China’s export control catalogue and restrict overseas access to Chinese AI model weights has closed without a formal announcement. The proposed tiered restriction framework - covering open-source filing requirements, security reviews, and a potential ban on overseas release of the most capable models - remains unresolved with no new timeline announced.