Is it safe to send a customer's crash dump?
What a minidump, a kernel dump and a full memory dump can contain, what to send instead of a full dump, and what DumpTriage keeps and for how long.
Short answer: It can be, if you send the smallest dump that shows the crash and the service says in writing where the dump is stored, who reads it, how long it is kept and whether it trains models. A minidump holds far less than a kernel or full memory dump, and any dump can hold fragments of what was in memory when the machine crashed. Get the customer's agreement before it leaves their network.
What each kind of dump can contain
| Dump | What it holds | Risk |
|---|---|---|
| Minidump (Windows) | The stop code, the list of loaded modules, the stack of the crashing thread and a small amount of other memory. | Lowest. Stacks can still hold a path with a user name or a string a program was working on. |
| Kernel memory dump (Windows) | The memory the kernel and drivers were using, but not the memory of ordinary user programs. | Higher. Network buffers, file-system cache and driver data can contain fragments of what the machine was doing. |
| Full memory dump or application dump with full memory | All of physical memory, or all of one process's memory. | Highest. Passwords, keys, tokens, documents and database rows can be in it. |
| Linux core or vmcore | The memory of one process (core), or of the kernel (vmcore, depending on the kdump filter). | As for the Windows equivalent. A core of a database or web server is a full application dump. |
How the Windows dump types differ, and how to choose one, is in how to analyze a Windows minidump; the Linux equivalents are in Linux core dumps with gdb and crash.
What to send instead of a full dump
- A minidump, when the stop code and the stack are enough. For most bug checks, such as 0x3B, 0xD1 or 0x133, they are.
- A dump from a test or staging machine that shows the same crash, when the production machine handles data you must not move.
- The debugger's own text output (
!analyze -v,kb), when you can read it first and decide what to share.
Send a kernel or full dump only when the smallest one did not settle the cause, and only with the customer's agreement.
What to ask any service before you send
- Where is the dump stored and where is it debugged, and which third parties touch it?
- Does an AI model see the dump itself, or only the debugger's text output, and is that output scrubbed first?
- Is the dump used to train or fine-tune any model, by the service or by its AI provider?
- How long are the dump and the transcripts kept, can you delete them yourself, and what remains afterwards?
- Will the service sign a non-disclosure agreement, and who can open the dump (an engineer, a contractor)?
What DumpTriage does with a dump
In the words of our AI and data page:
- No AI model sees your dump. The debugger runs on our own worker. A model only reads the debugger's text output, after we redact it.
- We don't train or fine-tune any model on your dumps, transcripts or reports, and our AI provider's terms say it won't either.
Dumps are stored in the EU and debugged in the EU. When a case goes to an engineer, they work with the dump itself; if you need a non-disclosure agreement first, ask us and we'll sign one.
Redaction is pattern-based and best-effort, not a guarantee: a secret in a shape we don't recognise can get through. That is why the AI and data page suggests keeping the most sensitive systems out of scope or sending minidumps rather than full memory dumps. DumpTriage isn't built for protected health information, and we don't sign HIPAA Business Associate Agreements.
What we keep, and for how long
- The dump file and the full debugger transcripts produced from it are deleted 30 days after upload by default, or after the retention period configured for your account (at most 90 days). You can delete a dump yourself at any time through the API.
- The report's findings, and for escalated cases the engineer's notes, are deleted with the dump.
- What remains of the dump's record is metadata only: its size, type, upload date, file hash and a keyed hash of the uploading machine's name, deleted 6 years after the dump.
- Private symbol files you upload are kept until you delete them.
The Privacy Policy has the full retention section, including backups and what happens when an account is closed, and the subprocessors list names every third party and its location.
Common questions
- Is it safe to send a crash dump to an analysis service?
- It depends on the dump and the service. A minidump holds far less than a kernel or full memory dump. Check where the dump is stored, who reads it, how long it is kept and whether it trains models, and send the smallest dump that still shows the crash.
- What is the smallest crash dump I can send?
- A minidump. It holds the stop code, the loaded modules and the stacks, not the whole of memory, and for most bug checks it is enough to name the faulting module. Send a kernel or full memory dump only when an engineer asks for one.
- How long does DumpTriage keep a dump?
- The dump file and the debugger transcripts made from it are deleted 30 days after upload by default, or after the retention period set for your account (at most 90 days). You can delete a dump yourself at any time through the API.
Ready to send a dump?
Upload a minidump to DumpTriage. It runs the debugger, reads the output and sends back the faulting module, the stack and a plain-language root cause (how the analysis works). Cases the automated passes can't settle go to an engineer. Free during the pilot.
Create an account