Investigators and legal teams who need to defend AI-assisted findings now have a governed way to run large language models over evidence. Sydney-based Nuix has announced the Neo Generative AI Framework, an orchestration layer that connects a customer's chosen LLM to data processed by the Nuix Neo platform.
The framework combines audit-grade telemetry with policy-enforced guardrails, designed to support evidentiary integrity and corporate usage controls.
It monitors and meters token usage to prevent runaway costs. It also manages the context window to ground answers in source data.
Customers keep the flexibility to choose their own LLM and provider. Nuix provides the governance framework that connects that model to its forensically processed data layer.
The Los Angeles County District Attorney's Office has used the framework to select AI models and integrate them into its Nuix Neo workflows.
"Being able to leverage the latest AI advancements while staying confined within our own environment is an accelerator," said Donn Hoffman, Chief Privacy Officer and Deputy District Attorney at the office.
Two capabilities sit on the framework. Nuix Neo BYO AI applies repeatable workflows to analyse every item in a dataset, rather than prioritising or sampling subsets.
It produces structured reports with links back to source items and citations.
The Nuix Neo AI Agent is a conversational investigation capability for interrogating data across text, metadata and images. It draws on Nuix's APIs and a curated knowledge graph.
The AI Agent is available to early adopters through the Nuix Responsible Innovation Framework program.
Nuix says outputs from any generative AI capability remain probabilistic. They must be verified against source material and are intended to support, not replace, the judgement of qualified investigators, reviewers and legal professionals.