Labs / Time machine / Datalab
Datalab
The subprocessor list as we last read it, on Nov 22, 2026.
Dates are when we read the list (UTC), not when Datalab changed it. Between two readings we don't know what it said.
tall: list changed · medium: document changed · short: read, no change · x: check failed · dotted: change dates only. Tap a tick or use the date box.
We record change dates, not every check, before Oct 11, 2026.
Read Nov 22, 2026 · 7 companies named as a subprocessor · stored document
| Name | Purpose | First read |
|---|---|---|
| Anthropic AI | LLM inference for extraction and agent features. | |
| AWS AI | S3 storage for batch run inputs and outputs; Amazon Bedrock as LLM inference. | |
| Cloudflare | DNS, CDN, edge, and R2 object storage. Document content transits Cloudflare's edge in transit; R2 stores consent-gated retained content and operational data. | |
| GCP AI | Hosting, storage, and model inference in the US and EU (EU in europe-west4, Netherlands), including LLM inference through Vertex AI, called directly for BAA and EU-residency customers. Processes and stores customer data, including document content, prompts, and outputs. | |
| Modal | Serverless GPU infrastructure for model inference. Processes customer document content transiently during processing. | |
| OpenAI AI | LLM inference through OpenRouter; may transiently process document content. Customer data is not used to train models. | |
| Openrouter AI | Routes LLM inference to Google, OpenAI and Anthropic. |
Built from the changes we published; changes we recorded without an alert (parser fixes, entries approved after review) don't appear. Where we later improved how we read a list, an earlier list may look different from the vendor's page at the time.
Gaps in our readings
No gaps since Oct 11, 2026.
All readings (46)
History starts Oct 8, 2026. Dates are UTC. Data CC BY 4.0. Shown as stored when this page was built.