Australia's AISI: AI Governance Gap in Multi-Org Systems

Australia's AI Safety Institute has identified a critical gap in current AI governance frameworks. All major frameworks assume one organization controls all AI agents in a system. The Institute's new report, released on August 9, 2026, is the first government publication to address the risks when this assumption breaks down. This happens when AI agents from different organizations interact. Supply chains and customer-facing AI already involve automated exchanges between various systems. An agent from one organization might interact with an agent from another, potentially with different or incompatible safety constraints. This creates a new problem called "cross-organizational agent interaction." Existing governance levers designed for a single organization don't work across organizational boundaries. As one expert puts it, "a collection of safe agents does not guarantee a safe collection of agents." This report highlights that a collection of governed agent systems does not make a governed cross-organizational system. This matters because the increasing use of AI across different organizations creates new risks that current frameworks don't address.