Secure explanation generation by large language models for sensitive search targets

A system, comprising:

• a computer processor;
• a large language model (LLM), executing on the computer processor;
• a small language model (SLM), executing on the computer processor;
• a topic analysis model, executing on the computer processor;
• and
• a secure explanation engine, executing on the computer processor, interfacing with the LLM, the SLM, and the topic analysis model, and configured for:
• generating, using the LLM, a plurality of match explanations for a plurality of search targets,
• clustering, using the topic analysis model, the plurality of match explanations to obtain a plurality of explanation clusters,
• generating, using the LLM, an explanation template for an explanation cluster of the plurality of explanation clusters to obtain a plurality of explanation templates,
• obtaining, by a matching model of the secure explanation engine, a target match for a search target received from a user application,
• generating, using the LLM, a corresponding match explanation for the target match,
• matching an encoding of the corresponding match explanation to a plurality of encoded templates corresponding to the plurality of explanation templates;
• responsive to the encoding matching an encoded template of the plurality of encoded templates, populating, by a small language model (SLM), the corresponding match explanation from the search target according to a format defined in the encoded template;
• transmitting the corresponding match explanation to a human-in-the-loop (HITL) portal;
• scanning, by a security checker employing a plurality of filters in the HITL portal, the corresponding match explanation to identify a security threat and to ensure a content in the corresponding match explanation passed to the user application for a particular user session is privileged exclusively to the particular user session;
• and
• returning the content to the user session.