IBM releases LLMON-enabled Granite models to combat prompt injection
In short: IBM has introduced LLMON (LLM Object Notation), an XML-like markup language that uses reserved special tokens to explicitly label components like instructions, data, and constraints. Two experimental models, Granite 4.1 3B and 8B LLMON-preview, are now available, demonstrating near-zero success rates for indirect prompt injection attacks.
This summary was generated automatically by AI from Hugging Face's publication. It is our own text, not a copy of the original — facts, figures and quotes belong to the source, linked above and below.
What changed?
- 1Introduced LLMON format for explicit instruction/data separation.
- 2Released experimental 3B and 8B models with LLMON-native training.
- 3Added support for execution blocks that bind instructions to specific data payloads.
- 4Provided conversion tools for existing JSON pipelines.
| Parameter | Before | Now |
|---|---|---|
| Architecture | Standard Granite 4.1 Instruct | LLMON-native structure with special-token delimiters |
| Context Handling | Flat text/JSON | Addressable blocks (instruction, data, requirement, exec) |
Why it matters
This approach addresses the brittleness of standard models when processing unstructured data, such as RAG retrieval results or tool outputs. By isolating malicious instructions hidden in data, it prevents the model from being hijacked while maintaining parity in general reasoning tasks.
What it means for AI agents and contact centers
For teams building voice agents, this architecture allows for safer handling of retrieved information. When testing, focus on whether the explicit 'exec' blocks improve reliability in complex RAG workflows and if the security gains in preventing injection justify the transition from standard JSON-based prompt structures.
🧪 Worth Testing
Testing is recommended to evaluate how this structural approach impacts agent reliability and security in production-like workflows.
Sources
- Hugging FaceOfficialPrimary sourceOriginal article →„LLMON: An LLM-Native Markup Language, now in Granite 4.1 3B and 8B“12 Nov 2026, 21:07
- Published by source
- 12 Nov 2026, 21:07
- Found by our system
- 12 Nov 2026, 21:12
- Summary generated
- 12 Nov 2026, 21:43
This article was written by AI from the original source. Facts, numbers and prices come from the source; missing values are marked “Not specified”. Legal notice, copyright and privacy