Every field is source-linked and dated.See the rubric behind the grades.

How we grade
Sovereign AI Registry
ExploreBlogGov accessCertsCountries

Footer

Sovereign AI Registry

The compliance registry for AI vendors. Data residency, training defaults, retention, subprocessors and EU AI Act posture — one row per vendor, product and deployment, every claim linked to its source.

Registry

  • Explore vendors
  • Deployment models
  • Categories
  • Countries

Compliance

  • Gov access exposure
  • EU AI Act roles
  • Certifications

Resources

  • FAQ
  • Methodology
Built with ShipMore·Build yours →

© 2026 Sovereign AI Registry. All rights reserved.

  1. Explore
  2. Whisper, self-hosted
WS

Whisper, self-hosted

Sovereignty grade A (95/100). You control where it runs; no training on your data; self-hosted weights. OpenAI, United States.

Category
meeting-ai
EU AI Act role
Deployer
Government access exposure
No non-EU government access
Deployments
Self-hosted weights
Countries
United States
Categories
meeting-ai

Details

Deployment
self-hosted-weights
Hq Country
United States
Hq City
San Francisco
Ownership
private
Founded
2022
Trust Center Url
github.com/openai/whisper
Training Default
never
Residency Options
wherever the operator runs it, fully offline operation is the normal case
Gov Access Exposure
none-eu
Eu Ai Act Role
deployer
DPA available
structurally-na
Zero Retention Available
Yes
Self Host Option
Yes
Retention Default
There is no service in the path. The model weights and the reference implementation are released under the MIT licence; audio is transcribed on the operator's own machine and nothing is transmitted. Retention is whatever the operator's own storage policy is.
Retention Exceptions
Whisper is a model, not a product: it has no meeting bot, no calendar integration, no sharing and no support contract. A team adopting it is taking on the integration work that Otter or Fireflies sells. Production deployments generally use a reimplementation such as faster-whisper (CTranslate2) rather than the reference repository, which changes the maintenance question but not the data path.
Subprocessors
none — nothing leaves the operator's infrastructure
Subprocessor Count
0
Transfer Mechanism
intra-eu
Eu Procurement Ready
conditional
Eu Procurement Reason
This is the architectural floor of the category and it is here to make the comparison honest: transcription does not require sending a recording to anyone. MIT-licensed weights, an MIT reference implementation, and mature reimplementations mean a meeting can be transcribed on a laptop or on a rack in the customer's own building, with no vendor in the path to disclose, no sub-processor to audit and no training question to ask. Every hosted row in this batch is paying for the product built on top of that: the bot that joins the call, the calendar integration, the summaries, the search, the support contract. The condition is that a buyer must supply all of it themselves. The score measures the data path and the data path here is unbeatable; it does not measure that you have just taken on an engineering project.
Training Evidence Url
github.com/openai/whisper
Training Claim Basis
inferred
Training Confidence
high
Residency Evidence Url
github.com/openai/whisper
Residency Claim Basis
inferred
Residency Confidence
high
Retention Evidence Url
raw.githubusercontent.com/openai/whisper/main/LICENSE
Retention Claim Basis
inferred
Retention Confidence
high
Subprocessors Evidence Url
github.com/openai/whisper
Subprocessors Claim Basis
inferred
Subprocessors Confidence
high
Residency Class
customer-controlled
Portability / Exit Path
open-weights-permissive
Subprocessor Jurisdiction
structurally-none
Underlying model providers
none — MIT-licensed weights run on the operator's own hardware
At a glance
Governance Grade
A
Governance Score
95
EC SOV-2 Legal & Jurisdictional (0-4)
3
EC SOV-3 Data & AI (0-4)
3
Website

Similar

Read AI · Sep 2026

RARead AI meeting assistant

Sovereignty grade F (24/100). No EU residency. Read AI, United States.

meeting-ai
Governance Grade
F
Fathom · Aug 2026

FAFathom AI notetaker

Sovereignty grade D (35/100). No EU residency. Fathom, United States.

meeting-ai
Governance Grade
D
Granola · Sep 2026

GAGranola AI notepad

Sovereignty grade D (30/100). No EU residency. Granola, United States.

meeting-ai
Governance Grade
D

More EU AI Act role

  • Deployer59
  • General-purpose AI (GPAI)10
  • provider9
  • GPAI with systemic risk5

More Government access exposure

  • US CLOUD Act44
  • No non-EU government access27
  • Mixed jurisdiction9
  • PRC National Intelligence Law2

More Deployments

  • Hosted API68
  • Self-hosted weights6
  • On-premise6
  • Private VPC2

More Countries

  • United States45
  • Germany9
  • France7
  • Switzerland5
  • Singapore3
  • Israel2
  • United Kingdom2
  • China2
  • Netherlands2

More Categories

  • Coding assistant15
  • meeting-ai13
  • Sovereign host13
  • Inference host13
  • SaaS-embedded11
  • Foundation model11
  • Cloud platform7