SD-RAG: A Prompt-Injection-Resilient Framework for Selective Disclosure in Retrieval-Augmented Generation
Aiman Al Masoud, Marco Arazzi, and Antonino Nocera · 2026
Read SD-RAG: A Prompt-Injection-Resilient Framework for Selective Disclosure in Retrieval-Augmented Generation by Aiman Al Masoud, Marco Arazzi, and Antonino Nocera (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2601-11199.txt.
Description
Aiman Al Masoud, Marco Arazzi, and Antonino Nocera tackle a safety flaw in search-based AI. Their 2026 preprint builds a shield for retrieval-augmented generation. Such systems pair a language model with search over big text stores. The pair can write with facts pulled fresh from the store.
The risk sits in what the search hands to the model. Most designs send all found text straight into the prompt. That can leak private or locked facts to the model and then to users. It also opens a path for prompt injection attacks. The team's frame shares facts on a need-to-know basis. It aims to block leaks while the model still writes good text.
Catalog record
| CREATOR | Aiman Al Masoud, Marco Arazzi, and Antonino Nocera |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 46 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2601-11199