Nuix announces Neo Generative AI framework for governed use of LLM – Unite.AI

Nuix on October 7, 2026 announced the Nuix Neo Generative AI (Gen AI) framework, a governed orchestration layer to run generative AI on organizations’ structured and unstructured data. Nuix said the framework addresses governance, trust and verifiability concerns that have limited the use of large language models (LLMs) and Gen AI in high-stakes legal, risk and investigative matters.

Guardrail, telemetry and model choice

According to the announcement, the framework combines audit-grade telemetry with policy-enforced guardrails to support evidence integrity and business use controls for LLMs. Nuix said it monitors and measures token usage to prevent runaway token costs, and manages the context window to ground responses and strengthen the reliability of AI results.

Under the framework, a client’s selected LLM is linked to business data that is forensically curated and defensible by the Nuix Neo platform. Customers retain the flexibility to choose the model and vendor that best fits their needs, strategic goals and appetite for innovation, while Nuix provides the governance layer that securely connects that model to the curated data intelligence layer provided by the platform’s patented data processing engine.

Nuix said the Nuix Neo platform draws on more than 25 years of research and development to provide forensic-grade processing of unstructured data, creating a layer of data intelligence that hundreds of high-profile, high-risk organizations rely on. The Gen AI Framework builds on that layer to operationalize each client’s selected LLM application.

BYO AI and the AI ​​agent

The framework supports two features based on its enterprise-grade infrastructure: Nuix Neo BYO AI and Nuix Neo AI Agent. Nuix said BYO AI applies repeatable workflows that systematically process each item in a dataset rather than prioritizing or sampling subsets, then produces structured reports with links to source items and citations, giving customers clarity and control over token usage.

Nuix said the framework allowed the Los Angeles County District Attorney’s Office to select the AI ​​models best suited to its business and integrate them into Nuix Neo workflows. A case study published by Nuix states that the bureau turned to Nuix Neo Investigations with Nuix Neo’s latest AI capabilities to integrate AI models into its workflows. Donn Hoffman, chief privacy officer and deputy district attorney at the office, said that with Nuix Neo’s latest AI his office can move quickly to implement targeted solutions to digital evidence challenges and investigative bottlenecks while meeting compliance needs, and described the ability to apply the latest AI advances within the office’s environment as an accelerator.

Nuix Neo AI Agent is a conversational inquiry capability built on the Gen AI Framework and available to Early Adopter participants through the Nuix Responsible Innovation Framework. Nuix said the agent allows analysts to query their data by reasoning about text, metadata and images, with access to Nuix APIs and context enhanced by a curated knowledge graph, and that it is fully integrated into Nuix Neo so analysts can follow any lead in a managed, controlled and secure manner.

Nuix said that, as with any Gen AI feature, the results remain probabilistic, must be verified against source material, and are intended to support rather than replace the judgment of trained investigators, analysts, auditors, and legal professionals. Organizations interested in the Responsible Innovation Framework program or seeing the Gen AI Framework in action can contact their Nuix representative or visit the Nuix Neo website.

Platform Engine and AI governance

On the Nuix Neo product page, the company describes the generative framework as the Connected AI category of the platform, which adds orchestration and guardrails and then connects to the customer’s chosen tools via MCP. It sits alongside Nuix AI, the company’s deterministic artificial intelligence and natural language processing that returns the same outputs for the same inputs, and integrated AI capabilities such as semantic search, transcription and facial recognition.

Nuix says the platform’s patented engine reads more than 1,000 file types at the binary level, including corrupt, encrypted and proprietary files, at up to 1.5 terabytes per hour and more than 20 terabytes per day, with performance subject to the customer’s environment and hardware. The company also says the platform records its audit trail at the time of processing rather than reconstructing it later, with data sovereignty and implementation choice, and that organizations facing high-risk data challenges have relied on Nuix for 25 years, including the investigation behind the Panama Papers.

In a Nuix blog post on AI governance, Christopher Stephenson, head of AI strategy and governance at Nuix, wrote that the BYO-AI framework supports connections to OpenAI, Anthropic Claude, Google Gemini, AWS Bedrock, and Azure AI Services, with audit trails recording the model name, timestamp, prompt, and response for each interaction. He wrote that customers can also deploy local models on-premise via Ollama or VLLM in fully air-gapped configurations where no data leaves the network, an option he says will be available today.

Stephenson wrote that Nuix’s AI governance and responsible use policy, adopted by the company’s board of directors with ongoing oversight by the Audit and Risk Committee, is informed by the EU AI Act, ISO/IEC 42001, the NIST AI Risk Management Framework and the Australian AI Ethics Framework. He wrote that the policy is reviewed at least annually and independently evaluated under the company’s ISO/IEC 42001 certification program, and that every AI system shipped by Nuix is ​​registered, rated by impact level, and approved through an escalating authority that culminates at the board level for maximum impact capabilities.