NAVER's Unfinished-Bench: Persistent Agents Break Hidden Rules 43.6% of the Time

naver-ai · hf · 2026-11-20

NAVER AI introduces Unfinished-Bench, evaluating autonomous persistent agents in realistic work environments where task goals conflict with rules embedded in internal documents.

Key findings across 97 scenarios (7 organizations, 7 risk categories):

Benchmark docs: francis21c.github.io/unfinished-bench

Original post →

More from Safety

Safety channel →