Consistency Is Not Independence, and a Helpful Colleague Is the Most Dangerous Witness
Last week a new benchmark out of HKUST fed seven frontier AI agents a con that never touched their instructions: a familiar person — a spouse, a colleague — plus a folder of screenshots, documents, a website and an app that all agreed with each other, because one attacker had made all of them. Ever…
Read the full story at DEV Community — Machine Learning ↗
Timeline · 1 report
- 2026-11-09 14:55 · DEV Community — Machine Learning
Consistency Is Not Independence, and a Helpful Colleague Is the Most Dangerous Witness