Published May 13, 2026 | Version 0.1-seed
Working paper Open

Confident Misalignment as an Adversarial Attack Surface

Authors/Creators

  • 1. Non Sequitur Publishing

Description

The dominant discourse on agentic-AI output errors treats them as reliability\

Files

confident-misalignment-attack-surface-v0-1-seed.pdf

Files (706.3 kB)

Additional details

Related works

Is identical to
Working paper: https://nonsequitur.tech/pubs/white-papers/confident-misalignment-attack-surface/ (URL)