NNeriant

Counsel AI

A clinical research and investigation assistant — under governance.

Counsel AI is built around a model broker. Authorised clinicians choose from approved LLM providers. Every prompt, context package and output is audited. Humans stay in the loop.

Model broker

Choose the model. Keep the governance.

Anthropic Claude
Reasoning, long-form clinical reading
Approved
Google Vertex AI
Multimodal, customer-hosted Google Cloud routing
Approved
AWS Bedrock
Region-locked, in-VPC inference for SaaS tenants
Approved
Meta Llama (self-hosted)
Private/sovereign deployments, EU operating mode
Approved

Controls

What 'governed' actually means.

Governed model broker

Authorised clinicians choose from approved LLMs per tenant policy. Region- and purpose-locked routing.

Human oversight

AI assists; it never autonomously diagnoses, prescribes, places orders, or releases patient communications.

Full audit

Prompt, context package, model selection, output, clinician review and final action — all audit logged.

EU operating mode

EU-only or customer-approved region/model routing aligned with EHDS, GDPR Article 9 and AI Act expectations.

Research workbench

Secondary use, under permit and oversight.

Research access flows through the consent and policy engine: lawful basis, permit workflow, de-identification or pseudonymisation, secure processing environment, audit.

  1. Step 1
    Cohort request
  2. Step 2
    Permit & lawful basis
  3. Step 3
    De-identify / pseudonymise
  4. Step 4
    Secure workspace
  5. Step 5
    Audit & export controls