Cursor makes auto-review the default, with 97% accuracy
Cursor has made auto-review the default for all new users, using a classifier subagent that reviews actions in context and decides whether to allow, block, or request approval, with evals showing 97% accuracy.
Score breakdown
Auto-review becoming the default means new Cursor users get automated action-level oversight out of the box, without needing to configure it manually.
- 01Auto-review is now the default for all new users of Cursor.
- 02A classifier subagent reviews each action in context before deciding to allow, block, or ask for approval.
- 03Cursor's evals show the classifier is 97% accurate.
Cursor has rolled out auto-review as the default setting for all new users. The system works by running a classifier subagent that examines each action in context before making a three-way decision: allow the action, block it, or ask the user for explicit approval.
According to Cursor's own evaluations, the classifier achieves 97% accuracy.
According to Cursor's own evaluations, the classifier achieves 97% accuracy. The post notes that most misses cluster near ambiguous edges, suggesting the model performs reliably on clear-cut cases while occasionally struggling with borderline situations.
Key facts
- 01Auto-review is now the default for all new users of Cursor.
- 02A classifier subagent reviews each action in context before deciding to allow, block, or ask for approval.
- 03Cursor's evals show the classifier is 97% accurate.
- 04Most accuracy misses occur near ambiguous edge cases.
Topics
Summary and scoring are generated automatically from the original article. We always link back to the publisher and never republish images or paywalled content. Last processed Jun 12, 2026 · 10:05 UTC. How this works →