# Queen's University Belfast study finds AI agents strain zero-trust security

> Original title: AI Agents Are Testing the Limits of Zero-Trust Security

- Source: IEEE Spectrum · AI
- Published: 2026-11-09T16:25:06.000Z
- Bad Influence: https://badinfluence.news/items/a6bhxfhm2on5djns99wsgi6ke
- Original: https://spectrum.ieee.org/zero-trust-agentic-ai-architecture

## Summary

Queen's University Belfast researchers argue that AI agent systems strain zero-trust security, since a single policy enforcement point is hard to scale across agentic interactions. In an in-house experiment, agentic orchestration cut the time for a de-authentication attack to as little as 2 to 3 seconds. The authors call for AI bills of materials or distributed ledger traceability to establish trust under asynchronous autonomy.