Vexa LogoVexa

Privacy-First Meeting Transcription for Financial Institutions: Why Banks Cannot Use Google Meet

DGDmitry Grankin
December 4, 2025

Financial institutions need meeting transcription for compliance, documentation, and to power AI workflows—but cannot use Google Meet, Microsoft Teams, or Zoom transcription due to privacy regulations and data sovereignty requirements. Board meetings, M&A discussions, regulatory calls, and trading floor conversations contain sensitive information that cannot be sent to external cloud services.

Self-hosted meeting transcription is the only viable solution for banks that need to comply with GDPR, MiFID II, Basel III, and other strict regulatory frameworks. Unlike cloud-based solutions, self-hosted transcription keeps all data within your infrastructure—ensuring sensitive conversations never leave your control.

The Vexa open-source meeting transcription API is specifically designed for financial institutions, with self-hosted deployment on Red Hat OpenShift, native integration with OpenShift AI, and proven deployments at institutions like the Central Bank of Austria.

The Problem: Why Banks Can't Use Cloud Transcription

Regulatory Compliance Requirements

GDPR (General Data Protection Regulation):

  • Article 25 requires data protection by design and by default
  • Article 44-49 restrict data transfers outside EU without proper safeguards
  • Cloud transcription services process audio on external servers (compliance risk)
  • Result: Banks cannot use Google Meet, Teams, or Zoom transcription for sensitive EU meetings

MiFID II (Markets in Financial Instruments Directive):

  • Requires recording and retention of client communications
  • Data must be stored securely with full audit trail
  • Third-party cloud services create compliance gaps
  • Result: Trading floor conversations cannot be transcribed by external services

Basel III:

  • Requires robust operational risk management
  • Data sovereignty and control are critical for risk mitigation
  • Cloud dependency creates operational risk
  • Result: Banks need self-hosted solutions for operational resilience

Data Sovereignty Concerns

The Challenge:

  • Board meetings discuss M&A strategy, earnings, executive compensation
  • Regulatory calls with central banks contain confidential supervisory information
  • Trading floor conversations include material non-public information (MNPI)
  • Legal and compliance discussions must remain confidential

The Risk:

  • Google Meet transcription sends audio to Google Cloud servers
  • Microsoft Teams transcription processes data on Azure
  • Zoom transcription uses third-party cloud infrastructure
  • Banks have zero control over where data is processed or stored

Real-World Example: Central Bank of Austria

The Central Bank of Austria (OeNB) selected Vexa for on-premise deployment specifically because Google Meet transcription was not acceptable for regulatory discussions. All transcription processing happens within their Red Hat OpenShift infrastructure, ensuring complete data sovereignty.

The Solution: Self-Hosted Meeting Transcription on Red Hat OpenShift

100% Self-Hosted Architecture

Deployment Model:

  • Runs entirely on your Red Hat OpenShift cluster
  • No data ever sent to external cloud services
  • All transcription processing happens within your infrastructure
  • Full audit trail and data lineage for compliance

Technical Stack:

┌─────────────────────────────────────────┐
│     Your Red Hat OpenShift Cluster      │
├─────────────────────────────────────────┤
│  Vexa API (Bot Management & Retrieval)  │
│  Transcription Engine (Self-Hosted ASR) │
│  PostgreSQL Database (Enterprise DB)    │
│  OpenShift AI Integration (RAG, LLMs)   │
└─────────────────────────────────────────┘
          ↓ (All data stays inside)
    Your Network Perimeter

OpenShift AI Integration: Powering Financial AI Workflows

The Opportunity: Banks invest heavily in OpenShift AI infrastructure for:

  • RAG (Retrieval-Augmented Generation) pipelines
  • LLM-powered customer service and compliance monitoring
  • Agentic AI for trading analysis and risk assessment
  • Real-time conversational intelligence

The Challenge:

  • Existing transcription solutions don't integrate with on-premise AI stacks
  • Banks need real-time transcript streaming to AI workflows
  • Cannot send sensitive meeting data to external transcription APIs

Vexa Solution:

  • Native OpenShift AI Integration – Real-time transcript streaming to your AI workflows
  • API-First Architecture – RESTful API for RAG pipelines, vector databases, agentic AI
  • Supported LLM Models – Mistral, DeepSeek, IBM Queen, Llama Stack
  • Real-Time Data Layer – Conversational intelligence feeds your AI use cases

Example Workflow:

# 1. Request Vexa bot to join meeting
POST /api/meetings/request
{
  "meeting_platform": "google_meet",
  "native_meeting_id": "abc-defg-hij",
  "language": "en"
}
 
# 2. Stream real-time transcripts to OpenShift AI
GET /api/meetings/{meeting_id}/transcript
# → Feed to RAG pipeline for compliance monitoring
# → Send to LLM for real-time summarization
# → Store in vector database for semantic search
 
# 3. Power AI workflows entirely on-premise
# → All data stays within your OpenShift cluster
# → No external API calls to cloud services
# → Full regulatory compliance

Market Validation: Which Banks Use OpenShift?

80% Market Share in On-Premise Banking

Red Hat OpenShift has 80% market share in on-premise container infrastructure for financial institutions. All major banks in Spain and Portugal use OpenShift:

Spanish Banks:

  • Banco Santander
  • BBVA
  • Banco Sabadell

Portuguese Banks:

  • Banco de Portugal
  • Novo Banco

Why OpenShift?

  • Enterprise-grade Kubernetes with 30+ open source components (ArgoCD, Tecton, Elasticsearch, Grafana)
  • Regulatory compliance certifications (SOC 2, ISO 27001)
  • On-premise deployment for data sovereignty
  • Native AI capabilities with OpenShift AI add-on

Why Vexa for Financial Institutions?

1. Regulatory Compliance by Design

GDPR Compliance:

  • All data processing happens within your EU infrastructure
  • No cross-border data transfers
  • Full data lineage and audit trail
  • Data protection by design (Article 25)

MiFID II Compliance:

  • Secure recording and retention of client communications
  • Complete audit trail for regulatory reporting
  • Self-hosted storage with no third-party dependencies

Basel III Compliance:

  • Operational resilience through self-hosted deployment
  • No cloud dependency or vendor lock-in
  • Full control over data and infrastructure

2. Enterprise-Grade Security

Architecture:

  • PostgreSQL database (standard enterprise database, no proprietary storage)
  • Containerized on Red Hat OpenShift (proven enterprise platform)
  • RESTful API with OAuth2 authentication
  • All traffic stays within your network perimeter

Security Features:

  • Self-hosted transcription engine (no external API calls)
  • Encrypted data at rest and in transit
  • Role-based access control (RBAC)
  • Full audit logging for compliance

3. Simple Integration with Existing Systems

API-First Design:

# Request bot to join meeting
curl -X POST https://your-openshift-cluster/api/meetings/request \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -d '{
    "meeting_platform": "google_meet",
    "native_meeting_id": "abc-defg-hij",
    "language": "en"
  }'
 
# Get real-time transcript
curl https://your-openshift-cluster/api/meetings/{meeting_id}/transcript \
  -H "Authorization: Bearer YOUR_API_KEY"

Integration Points:

  • RESTful API for bot management and transcript retrieval
  • Webhooks for real-time event notifications
  • PostgreSQL database for standard SQL queries
  • OpenShift AI for LLM and RAG workflows

4. Open Source: Zero Vendor Lock-In

Apache 2.0 License:

  • Full source code transparency (audit for security and compliance)
  • Modify and extend to fit your needs
  • No proprietary dependencies or vendor lock-in
  • Community-driven development

Why Open Source Matters for Banks:

  • Transparency: Audit exactly how your audio is processed
  • Control: Deploy and customize on your infrastructure
  • Cost: No per-minute pricing, scale on your own hardware
  • Security: Community review ensures robust security

Competitive Analysis: Why Not Google Meet, Teams, or Zoom?

Feature Google Meet Microsoft Teams Zoom Vexa
Privacy Cloud-based Cloud-based Cloud-based 100% Self-Hosted
Data Sovereignty ❌ ❌ ❌ ✅
GDPR Compliant ⚠️ Challenging ⚠️ Challenging ⚠️ Challenging ✅ By Design
MiFID II Compliant ❌ ❌ ❌ ✅
OpenShift Integration ❌ ❌ ❌ ✅ Native
API-First Limited Limited Limited ✅ Full RESTful API
Vendor Lock-In ✅ Google Ecosystem ✅ Microsoft Ecosystem ✅ Zoom Ecosystem ❌ Open Source
On-Premise AI ❌ ❌ ❌ ✅ OpenShift AI

Key Differentiators:

  • Google Meet: Cannot use for sensitive bank meetings (privacy risk)
  • Microsoft Teams: Data processed on Azure (compliance concerns)
  • Zoom: Third-party cloud infrastructure (data sovereignty issues)
  • Vexa: 100% self-hosted, native OpenShift integration, open source

Getting Started: Deploy Vexa on Red Hat OpenShift

Prerequisites

  1. Red Hat OpenShift cluster (4.12+)
  2. OpenShift AI add-on (optional, for AI workflows)
  3. PostgreSQL database (managed or self-hosted)
  4. Docker registry (for container images)

Deployment Steps

Step 1: Add Vexa to OpenShift

# Clone Vexa repository
git clone https://github.com/Vexa-ai/vexa.git
cd vexa
 
# Create OpenShift project
oc new-project vexa-transcription
 
# Deploy Vexa API and transcription engine
oc apply -f openshift/deployment.yaml

Step 2: Configure PostgreSQL Database

# Create PostgreSQL database
oc new-app postgresql-persistent \
  -e POSTGRESQL_USER=vexa \
  -e POSTGRESQL_PASSWORD=your-secure-password \
  -e POSTGRESQL_DATABASE=vexa_db
 
# Wait for database to be ready
oc wait --for=condition=ready pod -l name=postgresql --timeout=300s

Step 3: Deploy Vexa Services

# Set environment variables
oc set env deployment/vexa-api \
  DATABASE_URL=postgresql://vexa:password@postgresql:5432/vexa_db \
  API_KEY=your-secure-api-key
 
# Expose API endpoint
oc expose service vexa-api --hostname=vexa.your-bank.com

Step 4: Test Deployment

# Request bot to join meeting
curl -X POST https://vexa.your-bank.com/api/meetings/request \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -d '{
    "meeting_platform": "google_meet",
    "native_meeting_id": "test-meeting-id",
    "language": "en"
  }'
 
# Get bot status
curl https://vexa.your-bank.com/api/bots/status \
  -H "Authorization: Bearer YOUR_API_KEY"

OpenShift AI Integration (Optional)

Connect Vexa to OpenShift AI for Real-Time AI Workflows:

# Stream transcripts to OpenShift AI RAG pipeline
import requests
from openshift_ai import LLMPipeline
 
# Get real-time transcript
transcript = requests.get(
    "https://vexa.your-bank.com/api/meetings/{meeting_id}/transcript",
    headers={"Authorization": "Bearer YOUR_API_KEY"}
).json()
 
# Feed to OpenShift AI LLM (Mistral, DeepSeek, IBM Queen)
pipeline = LLMPipeline(model="mistral-7b")
summary = pipeline.summarize(transcript["text"])
 
# Store in vector database for compliance monitoring
vector_db.insert(embedding=pipeline.embed(transcript["text"]))

Red Hat Partnership: ISV Program & Co-Selling

ISV (Independent Software Vendor) Program

Vexa is a Red Hat Partner Connect approved ISV partner:

  • Certification: Vexa is certified for OpenShift compatibility with Central Bank Austria pilot
  • Support: Vexa provides full product support on Red Hat platforms
  • Validation: Customer warranty through Red Hat ISV program

Go-To-Market Strategy

Distribution:

  • Red Hat account executives introduce Vexa to OpenShift customers
  • Joint value proposition: "OpenShift AI + Vexa for Meeting Intelligence"
  • Reference architecture: "Vexa on OpenShift AI" demonstrates real-world financial services use case

Aligned Incentives:

  • More workloads on OpenShift = more cores = more revenue for Red Hat
  • Privacy-first solutions drive on-premise adoption
  • Banks require self-hosted transcription—perfect fit for OpenShift

Proof of Concept: 30-Day Pilot Program

Pilot Program Details

Duration: 30 days
Scope: 5-10 meetings with your team
Goal: Test privacy, compliance, and OpenShift AI integration

What You'll Get:

  1. Self-hosted deployment on your OpenShift cluster
  2. Technical support from Vexa engineering team
  3. OpenShift AI integration examples (RAG pipelines, LLM workflows)
  4. Compliance documentation for GDPR, MiFID II, Basel III

Success Criteria:

  • ✅ All transcription processing happens within your infrastructure
  • ✅ Zero data sent to external cloud services
  • ✅ Real-time transcript streaming to OpenShift AI workflows
  • ✅ Regulatory compliance requirements met

Timeline

  • Week 1: Deployment and configuration on OpenShift
  • Week 2: Test 5-10 meetings, verify transcript quality
  • Week 3: Integrate with OpenShift AI workflows (optional)
  • Week 4: Compliance review and production readiness

Case Study: Central Bank of Austria

The Challenge

Central Bank of Austria (OeNB) needed meeting transcription for:

  • Regulatory discussions with commercial banks
  • Internal policy meetings
  • Supervisory call recordings for compliance

Requirements:

  • ✅ 100% self-hosted (no external cloud services)
  • ✅ GDPR compliant (data sovereignty in Austria)
  • ✅ Red Hat OpenShift deployment
  • ✅ Integration with existing IT infrastructure

The Solution

Vexa Self-Hosted Deployment:

  • Deployed on OeNB's Red Hat OpenShift cluster
  • All transcription processing within Austrian borders
  • PostgreSQL database for secure storage
  • RESTful API for integration with existing systems

The Results

Privacy & Compliance:

  • Zero data sent to external services
  • Full GDPR compliance with data sovereignty
  • Audit trail for regulatory reporting

Technical Performance:

  • Real-time transcription with < 2 second latency
  • Multilingual support (German, English)
  • 99.9% uptime on OpenShift infrastructure

Business Value:

  • Eliminated dependency on Google Meet transcription
  • Enabled AI workflows for compliance monitoring
  • Full control over sensitive regulatory data

Pricing: Self-Hosted vs Cloud Transcription

Cloud Transcription Pricing (External Services)

Google Cloud Speech-to-Text:

  • $0.024/minute for standard recognition
  • $0.048/minute for enhanced models
  • Cost for 10,000 hours/month: ~$28,800/month

Microsoft Azure Speech Services:

  • $0.016/minute for standard recognition
  • $0.036/minute for custom models
  • Cost for 10,000 hours/month: ~$21,600/month

Compliance Risk:

  • ❌ Data sent to external cloud (GDPR, MiFID II violations)
  • ❌ No data sovereignty guarantees
  • ❌ No control over data processing

Vexa Self-Hosted Pricing

Deployment Model:

  • One-time setup on your OpenShift cluster
  • No per-minute pricing
  • Scale on your own hardware
  • Cost for 10,000 hours/month: Infrastructure costs only (~$2,000-$5,000/month for compute)

Compliance Benefits:

  • ✅ 100% data sovereignty
  • ✅ GDPR, MiFID II, Basel III compliant by design
  • ✅ Full control over data processing
  • ✅ Zero vendor lock-in (open source)

ROI Calculation:

Cloud Transcription: $28,800/month = $345,600/year
Vexa Self-Hosted: $5,000/month = $60,000/year
Annual Savings: $285,600 (83% cost reduction)
+ Regulatory compliance benefits (priceless)

Technical Architecture Deep Dive

System Components

Vexa API Server:

  • RESTful API for bot management and transcript retrieval
  • OAuth2 authentication with role-based access control
  • Webhook support for real-time event notifications
  • Deployed as OpenShift pod with auto-scaling

Transcription Engine:

  • Self-hosted ASR (Automatic Speech Recognition)
  • Supports Whisper, Faster-Whisper, and custom models
  • GPU acceleration (optional, NVIDIA CUDA support)
  • Real-time streaming transcription

PostgreSQL Database:

  • Stores meeting metadata, transcripts, and audit logs
  • Standard SQL queries for compliance reporting
  • Encrypted at rest with OpenShift storage encryption
  • Automatic backups with OpenShift persistent volumes

OpenShift AI Integration:

  • LLM inference (Mistral, DeepSeek, IBM Queen)
  • RAG pipelines with vector databases
  • Agentic AI with Llama Stack
  • VLLM for model serving

Data Flow

┌───────────────────┐
│  Google Meet/Zoom │
│  (Meeting Audio)  │
└─────────┬─────────┘
          │
          ▼
┌───────────────────┐
│   Vexa Bot        │ ← Request via API
│   (Self-Hosted)   │
└─────────┬─────────┘
          │
          ▼
┌───────────────────────────────┐
│  Transcription Engine         │
│  (OpenShift Pod)              │
│  - Audio → Text               │
│  - Real-Time Processing       │
└─────────┬─────────────────────┘
          │
          ▼
┌───────────────────────────────┐
│  PostgreSQL Database          │
│  (OpenShift Persistent Volume)│
│  - Store Transcripts          │
│  - Audit Logs                 │
└─────────┬─────────────────────┘
          │
          ▼
┌───────────────────────────────┐
│  OpenShift AI (Optional)      │
│  - RAG Pipelines              │
│  - LLM Summarization          │
│  - Compliance Monitoring      │
└───────────────────────────────┘
     ↓
All data stays within your OpenShift cluster

Frequently Asked Questions

Q: Can Vexa transcribe languages other than English?

A: Yes! Vexa supports 70+ languages including:

  • European: Spanish, Portuguese, French, German, Italian
  • Asian: Chinese (Mandarin, Cantonese), Japanese, Korean
  • Other: Arabic, Hebrew, Hindi, Russian, and more

Self-hosted deployment allows you to choose transcription models optimized for your languages.

Q: What meeting platforms does Vexa support?

A: Vexa supports:

  • ✅ Google Meet
  • ✅ Microsoft Teams
  • ✅ Zoom
  • ✅ Webex (coming soon)

The bot joins meetings as a participant and transcribes in real-time.

Q: How does Vexa compare to Recall.ai?

A:

Feature Recall.ai Vexa
Deployment Cloud-only Self-hosted
Privacy Data sent to Recall cloud 100% on-premise
Open Source ❌ Proprietary ✅ Apache 2.0
OpenShift ❌ ✅ Native
Compliance ⚠️ Challenging for banks ✅ By design

Vexa is the self-hosted, open-source alternative to Recall.ai for financial institutions.

Q: What are the system requirements for self-hosting?

A:

  • CPU: 4 cores minimum (8 cores recommended)
  • RAM: 8GB minimum (16GB recommended)
  • Storage: 50GB for database and models
  • GPU: Optional (NVIDIA CUDA for faster transcription)
  • Platform: Red Hat OpenShift 4.12+

Q: How long does deployment take?

A:

  • Initial setup: 1-2 hours (clone repo, configure OpenShift)
  • Testing: 1-2 days (verify transcription quality, compliance)
  • Production: 1 week (full integration with OpenShift AI, security review)

Q: What support is available?

A:

  • Community Support: GitHub issues, Discord community
  • Enterprise Support: Direct engineering support, SLA guarantees
  • Red Hat ISV Support: Technical guidance for OpenShift certification

Next Steps: Get Started with Vexa

1. Schedule a Demo

See Vexa in action with your OpenShift environment:

  • Live demo: Request bot, view real-time transcription
  • OpenShift AI integration: See transcript streaming to LLM workflows
  • Compliance review: Discuss GDPR, MiFID II, Basel III requirements

📧 Contact: dmitry@vexa.ai
🌐 Website: vexa.ai
💻 GitHub: github.com/Vexa-ai/vexa

2. Start 30-Day Pilot

Deploy Vexa on your OpenShift cluster:

  • Self-hosted deployment with technical support
  • Test 5-10 meetings with your team
  • Integrate with OpenShift AI workflows
  • Compliance documentation for regulatory review

3. Join the Community

Conclusion: Self-Hosted Meeting Transcription is the Only Option for Banks

Financial institutions cannot use Google Meet, Microsoft Teams, or Zoom transcription for sensitive meetings due to privacy regulations, data sovereignty requirements, and compliance risks. Cloud-based transcription services send audio to external servers, creating GDPR, MiFID II, and Basel III violations.

Self-hosted meeting transcription on Red Hat OpenShift is the only viable solution for banks that need:

  • ✅ 100% data sovereignty (all data stays within your infrastructure)
  • ✅ Regulatory compliance (GDPR, MiFID II, Basel III by design)
  • ✅ OpenShift AI integration (power RAG pipelines, LLM workflows, agentic AI)
  • ✅ Zero vendor lock-in (open source Apache 2.0 license)

Vexa is the open-source meeting transcription platform built specifically for financial institutions, with proven deployments at the Central Bank of Austria and Red Hat Partner Connect ISV certification.

Get started today: Deploy Vexa on your OpenShift cluster in 1-2 hours, test with 5-10 meetings, and power AI workflows entirely on-premise.

📧 Contact: dmitry@vexa.ai
🌐 Website: vexa.ai
💻 GitHub: github.com/Vexa-ai/vexa

About the Author

Dmitry Grankin is the CEO and founder of Vexa, an open-source meeting transcription platform for enterprise. Based in Portugal, Dmitry works with financial institutions across Europe to implement privacy-first meeting intelligence solutions. Connect with Dmitry on LinkedIn.

Don't want to host it yourself?
Try Vexa Cloud for hassle-free meeting transcription with the same powerful API.
Get Started FreeView Pricing