SAF-M-7: Content Rendering Parity

June 1, 2026 ยท View on GitHub

Overview

Mitigation ID: SAF-M-7
Category: UI Security
Effectiveness: Medium-High
Implementation Complexity: Low
First Published: 2025-01-03

Description

Content Rendering Parity ensures that what users see in the UI exactly matches what is sent to the LLM for all types of content (tool descriptions, tool outputs, error messages, and other data). This prevents attacks that exploit differences between displayed and processed content, including hidden instructions in tool outputs or visual deception techniques.

Mitigates

Technical Implementation

[TO BE COMPLETED]

References

  • SAF-M-8: Visual Validation
  • SAF-M-4: Unicode Sanitization and Filtering

Version History

VersionDateChangesAuthor
0.12025-01-03Initial stubFrederick Kautz
0.22025-01-09Generalized to cover all content types, not just descriptionsFrederick Kautz