Subprocessors
Third parties who process data on our behalf to deliver the DumpTriage service, and what they see. Referenced from our Privacy Policy.
Last updated 2026-11-06
| Subprocessor | Purpose | What they receive | Location |
|---|---|---|---|
| OVH SAS (OVHcloud) | Hosting for the DumpTriage API, database, database backups and dump storage, since 2026-10-15 | Uploaded dumps and private symbol files (encrypted at rest), account and contact details, analysis reports, debugger transcripts, escalation notes and case summaries | Gravelines, France |
| Hetzner Online GmbH | Hosted the DumpTriage API, database and dump storage until 2026-10-15. The old server is switched off and kept only as a fallback until we delete it, planned for 2026-10-22. | A copy of what the service held on 2026-10-15: the same categories as OVH above | Nuremberg, Germany |
| Google LLC (Google Cloud) | Runs the Windows and Linux debuggers that analyse each dump, each on its own worker VM created for the job and deleted afterwards | The raw, decrypted dump file and any private symbol files you've uploaded, and the debugger's output. They're deleted from the worker when the analysis ends; if that fails, they're deleted before the worker's next job starts. | Eemshaven, Netherlands (europe-west4) |
| Google LLC (Google Cloud Vertex AI) | AI-assisted Tier 1/2 crash-dump analysis with Google's Gemini models (see How DumpTriage uses AI) | Redacted debugger command transcripts (not the raw dump file), plus notes from similar past cases (see our Privacy Policy). Google's terms for Vertex AI say it won't use them to train or fine-tune its models without our permission. | Google's EU multi-region endpoint, which keeps the request in the EU. |
| Anthropic, PBC (Claude and Claude Code) | The AI assistant our engineers use when they investigate a case or a problem with an analysis, and when we test AI models. It isn't part of the automated Tier 1/2 analysis, so it doesn't receive every dump's data. | The debugger and forensic tool output, transcripts and report text an engineer works with. This output isn't always redacted first, so it can include the names of processes, drivers and files, and user and folder names. Not the raw dump file. | United States |
| Resend | Sending account verification, upload confirmation and failure emails to customers, and alerts to our staff | Your contact email address and organization name, dump identifiers, and the text of the email. Staff alerts about escalated cases include crash details such as the faulting module and stack frames. Resend keeps sent emails and their logs for 30 days. | United States |
| Sand Dune Mail Ltd (SMTP2GO) | Standby for sending the same emails when Resend can't deliver them. It is not used while Resend works. | The same as Resend above: your contact email address and organization name, dump identifiers, and the text of the email, including crash details in alerts to our staff. SMTP2GO keeps the headers of sent emails for 35 days and, by its own policy, doesn't store message content; it may keep one email in a thousand for 35 days when it suspects abuse. | Our account is hosted in the EU: SMTP2GO sends, stores and processes our mail in its EU data centres (Amsterdam). The company itself is in Christchurch, New Zealand. |
| Linear Orbit, Inc. (Linear) | Tracking the cases our engineers work on: one issue per dump escalated to a human, including dumps you asked us to review | Your organization name and an internal customer identifier, dump and job identifiers, the dump type, and the reason the dump was escalated, which includes crash details such as the bugcheck or exception code, the faulting module and the top of the crash stack. No dump files, transcripts or contact details. | United States |
| Slack Technologies, LLC (Slack) | Notifying our engineers in a private channel when a case is opened or updated in Linear | What Linear's notifications carry: the case title (your organization name and the dump type) and details from the case, which can include crash details such as the faulting module. No dump files, transcripts or contact details. | United States |
| Grafana Labs (Grafana Cloud) | Monitoring and alerting for the DumpTriage service | Operational telemetry: request paths without their query strings, timings and status codes, internal dump and job identifiers, model names, queue sizes, and error messages from failed requests and jobs. No dump files, transcripts, reports or contact details. Grafana Cloud keeps traces and logs for up to 30 days and metrics for up to 13 months. | London, United Kingdom |
| Microsoft (public symbol server) | Symbol resolution during analysis | Module names and PDB identifiers, not dump contents | United States |
| Public Linux debug-symbol servers: the debuginfod servers of Ubuntu, Fedora, CentOS, elfutils and Debian, and the distributions' package archives | Symbol resolution during the analysis of Linux dumps | Build IDs of the kernel and the libraries in the dump, and the names and versions of the packages that carry their debug symbols. Not dump contents. | Wherever each project hosts its server |
| Porkbun LLC | Hosting for this website (not the DumpTriage service) | Standard web request data from site visitors, such as IP address and pages requested | United States |
Where dumps are processed. Dumps are stored in the EU and the debugger runs in the EU. The AI analysis runs on Google's EU endpoint, so the dump, the debugger and the automated AI analysis all stay in the EU. That is the default for every account; setting the EU data region through the API (PUT /api/v1/customers/me/data-region) makes it a requirement for yours. Resend, Linear and Slack receive the notification and case details listed for them in the United States, and Anthropic receives what's listed for it there when an engineer uses it. Microsoft's symbol server there, and the public Linux symbol servers wherever they're hosted, receive module and symbol identifiers.
Previously. Until 2026-09-24, the automated AI-assisted analysis ran on Anthropic, PBC (United States), which received the same redacted transcripts and case notes. From then until 2026-11-02 this page said nothing was sent to Anthropic. That left out our engineers' use of Claude, listed above. From 2026-10-13 to 2026-10-15 this page listed MiniMax (Nanonoble Pte. Ltd.) as the AI provider while Google's AI service was paused. Our records show every analysis in that period still ran on Google's Vertex AI, and nothing was sent to MiniMax. Until 2026-10-15 the service was hosted by Hetzner in Germany (see above). Before 2026-10-14 the debugger mostly ran in Iowa, United States (us-central1). Until 2026-10-15 the AI analysis ran on Google's global endpoint unless the account had set the EU data region.
This list covers every third party that receives customer data today. Our software can also connect to other services (for example issue trackers or chat tools for our own staff), but none of them receive customer data unless they're listed here. We update this page before any new service starts receiving customer data, and we notify active customers of material changes to it, consistent with our data processing terms.