Risks and Controls for Multi-Agent Systems: an analytical framework for deployment of AI agents across organisational boundaries

The report introduces an analytical framework for assessing risks that arise when AI agents interact, especially as those interactions cross organisational boundaries. It defines three deployment tiers—singular governance, federated governance, and open environments—each with distinct governance requirements. For each tier, the report examines risk factors, failure modes, and controls, identifying who can apply them and highlighting gaps that require collective action.

Machine-generated by The Flow from the publisher's headline and feed description — not written or checked by a human. The full article lives at arXiv AI.

• agents