# Preventing AI agents from going rogue with a new measurement approach

Security experts Bruce Schneier and Barath Raghavan argue that preventing AI agents from acting unpredictably requires a novel way of measuring their behavior. They propose that such measurements are essential for ensuring safe and accountable AI as these systems become more autonomous. The article highlights the growing challenge of maintaining control over advanced AI agents.

**Importance:** 3/5

## Sources

### Business
- [The Guardian Biz](https://www.theguardian.com/commentisfree/2026/jul/28/rogue-ai-agent-instructions) — Tue, 28 Jul 2026 10:00:02 GMT