An AI assistant that reads your portfolio, drafts your trade notes, or summarizes your research feels like leverage. Sometimes it is. But every convenience like that runs on a question most people never ask: where does the data go once the model has seen it, and who else might see it later?

The trade you did not know you were making

Traders guard their edge — a research angle, a screening method, a position before it is public — the same way they guard a password. Yet many of the same people paste that exact information into a general-purpose AI tool without asking what happens to the prompt afterward. If a provider retains inputs for training or logging, your research strategy is no longer only yours. You have traded a piece of your edge for a bit of speed, without pricing that trade at all.

Questions worth asking about any AI tool you use for research

• Is data retained after the session ends, and for how long?
• Is your input used to train future models, by default or only by opt-in?
• Can you verify the answer, or only take the provider’s word for it?
• Does the tool run on infrastructure controlled by a single company, or is there any independent way to confirm what actually happened to your data?

The same sovereignty logic, applied to models

This is the same problem self-custody solves for coins, applied to information instead of assets. A hardware wallet keeps your keys out of a custodian’s hands. The equivalent move for AI is choosing tools with zero data retention, verifiable behavior, and no single provider sitting as an unaccountable middleman between your research and your results.

That is the thinking behind what we call the Secure Shell Ledger — frontier-quality open models, run with zero data retention and a verifiable record of what actually ran, without funneling your prompts through one custodian’s black box. Cold storage for your coins; a Secure Shell Ledger for your models. We wrote the fuller case for it in Secure Shell Ledger: Sovereign AI.

A practical filter, before you paste anything in

• Would you say this out loud in a room full of competitors? If not, think twice before typing it into a tool whose retention policy you have not actually read.
• Is the convenience saving you real time, or just saving you the friction of organizing your own notes? Those are different kinds of value, and only one of them is worth trading exposure for.
• Could you get most of the same benefit from a tool that keeps nothing, rather than one that keeps everything by default? Increasingly, the answer is yes.

Convenience is not free. It is priced in exposure you often cannot see until it has already cost you. Ask the retention question before you paste in the research, not after.

Not financial advice.

Want to set up a research workflow that keeps your edge yours? Bring it to a Builder Session.