- Deployment
- hosted
- Hq Country
- Netherlands
- Hq City
- Amsterdam
- Ownership
- private
- Founded
- 2017
- Training Default
- never
- Residency Options
- Western Europe
- Gov Access Exposure
- none-eu
- Eu Ai Act Role
- provider
- Certs
- ISO 27001, ISO 9001
- DPA available
- yes
- Retention Default
- "All data is securely stored in Western Europe." The audio and the generated transcript are explicitly never used to train the language models, and the privacy policy extends that to third parties: "Third parties do not store data or use data for training purposes."
- Retention Exceptions
- The distinctive exposure here is human, not machine. For all languages Amberscript uses transcribers who see the audio file and the transcript in order to correct it. They have signed non-disclosure and data-processing agreements, they cannot download audio or transcripts because the data stays locked inside the platform, and Amberscript states all of this on the page rather than leaving it to be discovered. It is a materially different risk from a model endpoint and it is the reason this row is worth reading next to the fully automated ones.
- Subprocessors
- human transcribers under NDA and DPA, no download rights, Google (Workspace SSO and analytics, under a signed agreement)
- Subprocessor Count
- 2
- Transfer Mechanism
- intra-eu
- Eu Procurement Ready
- pass
- Eu Procurement Reason
- A Dutch vendor storing everything in Western Europe, certified to ISO 27001 and ISO 9001, carrying the TPN badge for content security, and stating flatly that neither it nor its third parties train on customer audio. It is also the most honest row in the batch about what actually happens to a recording: a person reads it. The reviewed tiers put a human transcriber in front of the audio and the transcript, under NDA and DPA and without download rights, and the privacy policy says so instead of describing them as an unnamed category of service provider. For a buyer whose recordings are sensitive, that disclosure is worth more than a vendor promising full automation and naming no one.
- Training Claim Basis
- stated
- Training Confidence
- high
- Residency Claim Basis
- stated
- Residency Confidence
- medium
- Retention Claim Basis
- stated
- Retention Confidence
- medium
- Subprocessors Claim Basis
- stated
- Subprocessors Confidence
- medium
- Residency Class
- eu-only
- Portability / Exit Path
- proprietary-only
- Subprocessor Jurisdiction
- mixed
- Underlying model providers
- Amberscript's own speech models, human transcribers for the reviewed service tiers
