AI-Assisted Transcription Drafting (Internal LLM Operations)

The contract outlines the development and operation of an internal large language model system designed to generate initial transcription drafts from audio recordings while maintaining strict data security protocols. The system must be fully contained within secure internal infrastructure to ensure that no sensitive audio content is ever transmitted to or processed by external AI platforms, eliminating any risk of data exposure or unauthorized access. This requirement underscores the need for end-to-end control over data pipelines, model training, and inference environments, with all operations confined to approved, secure environments. The work is classified as a subcontract under NAICS code 541512 for other computer-related services and is managed by the Judicial Branch through its Aousc-procurement Management Division. The place of performance is designated as Washington, DC, with a zip code of 20544, indicating the primary operational site is likely within a federal judicial facility. Responses to this solicitation are due by May 18, 2026, at 6:00 PM Eastern Time, and while no set-aside type is specified, the project’s nature demands expertise in secure AI deployment, audio-to-text processing, and compliance with federal data handling standards for sensitive judicial materials.