Your Will Is on OpenAI’s Servers. Here’s What That Means.
Someone uploaded their will to ChatGPT last week
They wanted to know whether the executor clause covered a specific scenario. Reasonable question. ChatGPT gave a reasonable answer. But here’s what also happened: OpenAI’s servers now have a copy of a document containing the full legal names of every family member, their home address, the executor’s contact details, a description of every asset, and the names of every beneficiary.
That’s not hypothetical. People upload wills, insurance policies, tax returns, and medical records to ChatGPT and Gemini every day. The AI works. The privacy implications are the part nobody thinks about until it’s too late.
What’s actually in a will
A will is one of the most information-dense documents a family owns. A typical Canadian will contains:
- Full legal names of the testator and all beneficiaries
- Home address and property descriptions
- Names and contact details of executors and trustees
- Detailed asset descriptions — bank accounts, property, investments, vehicles
- Guardianship designations for minor children
- Specific bequests with dollar amounts or asset descriptions
When you paste that into ChatGPT and ask a question, every one of those details is sent to OpenAI’s servers in the United States.
What happens to your data on OpenAI’s servers
OpenAI’s privacy policy is clear on several points, and genuinely ambiguous on others. Here’s what we know:
What’s confirmed
- By default, conversations and uploaded files may be used to train future models. You can opt out in settings, but the default is opt-in.
- Data is processed on US servers, subject to the US CLOUD Act — meaning US authorities can compel access regardless of your citizenship.
- Even with chat history disabled, OpenAI retains conversations for up to 30 days for abuse monitoring.
What’s less clear
- Whether opting out of training is retroactive — data submitted before you found the toggle may already be in a training dataset.
- How data is handled by subprocessors and infrastructure partners.
- What happens to data if OpenAI is acquired, restructured, or compelled by a court order.
Google’s Gemini follows a similar pattern. Training on user data is the default. Opt-out exists but requires deliberate action.
Why this matters more for wills than for meeting notes
If someone accesses your uploaded meeting notes, the worst case is embarrassment. If someone accesses your will, they know your family’s complete financial picture, who inherits what, who’s responsible for your children, and where you live.
This isn’t about whether OpenAI is irresponsible — it’s about whether a general-purpose AI chatbot is the right tool for documents this sensitive. The answer, to us, is clearly no.
How Archevi handles the same question differently
You can upload your will to Archevi and ask the exact same question about the executor clause. The difference is what happens between your question and the AI’s answer.
1. You ask: “Does the executor clause cover this scenario?”
2. Archevi detects entities: your name, your spouse’s name, your address, the executor’s name — all replaced with realistic surrogates.
3. The AI receives: a query about “James Chen” at “88 Oak Avenue, Halifax” — not your real family.
4. The AI processes the surrogates, finds the relevant clause, generates an answer.
5. You see the answer with your real names restored. The AI never knew them.
Your will’s content — the actual names, addresses, and asset details — never left our Canadian server. The AI only ever saw a fictional family.
The hard blocks
Some data is too sensitive even for surrogates. If your will mentions a Social Insurance Number, a bank account number, or a credit card number, Archevi doesn’t anonymize it — it blocks the query entirely. You’ll see a clear message explaining what was detected and why the query was stopped.
ChatGPT has no equivalent protection. Paste a SIN into the chat and it processes it without hesitation.
What this costs you
Archevi adds 1–2 seconds of latency to anonymize your documents. We think that’s worth it. You get the same answers about your family’s documents — with the guarantee that the AI never saw the real thing.
Your will stays on a server in Toronto, under Canadian privacy law. Not on a server in Virginia, under the CLOUD Act.
Try it with your most sensitive document
Sign up free. Upload the document you’d never paste into ChatGPT. Ask it a question. Then check the privacy vault to see exactly what the AI received — surrogate names, surrogate addresses, and nothing real.
That’s the difference between a tool that works and a tool you can trust.
