SecGPT

by Quantana

Private generative AI for organisations whose data can’t leave the building

BankingInsurancePharmaHealthcare

Your most valuable knowledge is the data you can’t show to AI

SecGPT brings the AI to your data, not your data to the AI

Your premisesYour servers, your documents, your model

Runs on your hardware

The whole platform is installed on-premises, inside your own network.

Works with no internet

Zero internet connectivity required, and no calls to outside services.

No cloud bills

Full offline, autonomous operation with no per-query cloud costs.

Knows your documents

Answers come from your own policies, procedures and files, with sources cited.

Shows its reasoning

Every reasoning step is visible, so answers can be checked and audited.

Learns your organisation

User feedback corrects errors and reveals gaps in your content.

Built for industries where a leak is not an option

Reads everything. Forgets nothing. Tells no one.

256K tokens of context

Whole documents, not snippets

Reason over entire manuals or multi-document datasets in one session.

140+ languages

One platform for every team

Translate language to language, with your industry’s terminology preserved.

Personal to each user

Persistent profiles adapt to each person’s progress, strengths and gaps.

Improves with use

Feedback loops, error correction memory, gap detection and document generation.

Auditable answers

Chain-of-thought reasoning helps prevent hallucination and keeps answers accurate.

Agents that stay inside

Run local scripts against internal databases, with no external connectivity.

From document to cited answer, entirely inside your network

Your documentsPDF, DOCX, TXT
Private ingestionParsing and vector encoding
Encrypted vector storeA local database, never hosted
SecGPT modelInference and response generation
Web and chat interfaceDelivered to authorised users
Admin dashboardUsage monitoring and role management
Secure storageYour private document repository
Network isolationA cryptographic boundary around it all

Every answer links back to the source document it came from.

The full stack, under your control

Security you can walk your auditor through

Air-gapped by design

Cryptographically isolated deployment with enforced zero outbound data flow.

Keys bound to hardware Our USP

Each node is tethered physically, so it can’t be cloned. Post-quantum cryptography enabled.

Sealed inference and memory

No data exits or replicates beyond the authorised node.

Model lineage

A unique cryptographic key per deployment supports model immutability and audit compliance.

Policy-led governance

A framework for secure, policy-aligned evolution of AI behaviour.

Access and audit

Role-based access, SSO, content filtering, audit logging and monitoring.

Post-quantum ready by design. Not a retrofit later.

Conventional application

Applications built today may run for 10 to 20 years, while conventional public-key cryptography may become vulnerable to quantum attacks.

  • Legacy cryptography
  • Harvest now, decrypt later risk
  • Expensive retrofit
  • Application disruption

SecGPT

Post-quantum readiness is built into the foundation. Built for records you must keep for decades.

  • PQC-ready architecture
  • Crypto agility
  • Continuous evolution
  • Quantum-ready future

Useful to everyone who touches regulated work

Compliance training that proves itself

Run certifications and mandatory assessments in a secure, fair and credible environment.

  • Browser and application monitoring

    Detects unauthorised navigation, multiple windows and prohibited applications.

  • Multiple person detection

    Identifies if more than one person is in the frame.

  • Prohibited device detection

    Flags mobile phones, secondary screens or unauthorised devices.

  • Identity verification

    Confirms the right person is taking the exam, through live facial verification.

  • AI-powered anomaly detection

    Detects suspicious behaviour patterns in real time.

  • Dashboard and alerts

    Real-time alerts and session insights for quick intervention.

Private doesn’t have to mean slow

≤180msinference latency
256Ktokens of context in one session
300concurrent users in the reference configuration
20 TB+knowledge base capacity

Network dependency: none. Full offline autonomous operation, zero cloud costs.

Live in twelve weeks

Install and secure
Ingest your knowledge
Tune and test
Go live
Weeks 1–3

Install and secure

Hardware installation, environment verification, storage integration.

Weeks 3–6

Ingest your knowledge

Your documents are embedded on-site, up to 10TB in the initial load.

Weeks 6–9

Tune and test

Reasoning is tuned to your content, then proven in user acceptance testing.

Weeks 10–12

Go live

Full deployment, admin and user training, documentation handover, support activation.

Everything arrives ready to run

Hardware

  • Dual AMD EPYC server
  • Four NVIDIA RTX Pro 6000 GPUs96GB each
  • 512GB DDR5 ECC memory
  • Enterprise SSD storage in RAID 5
  • Firewall, switches and rack

Software

  • SecGPT platform and web interface
  • Document intelligence pipelineWith retraining support
  • Admin dashboard
  • Server and user licences

People

  • Implementation and user onboarding
  • Admin and technical staff training
  • Three years of onsite supportMonitoring and security patching
  • New document ingestion services
  • Manuals, documentation and formal handover

Bring AI inside your walls

120+projects delivered
100+clients, including Fortune 500 companies
ISO 27001certified

Talk to Quantana

SecGPT by Quantana. Offices in Australia, the US and India.