Fifteen questions to ask about your AI
An AI that remembers you can be genuinely helpful. It also builds a picture of your life. These questions help you find out who else can see that picture, what it is used for, and whether you can take it with you or end it. You don't need to understand cryptography to notice when a company can't say who can read your information.
V Can the company read it?
Could anyone at the company read what my AI knows about me, even in an emergency?
Encryption only protects you from whoever does not hold the key. If the company holds it, the protection is a policy, not a property.
Warning sign: The answer is about rules and good intentions rather than who holds the keys.
When my AI uses a tool or the internet, where can my information go?
Sensitive information can escape through logs, tools, telemetry and fallback services even when the primary storage is encrypted.
Warning sign: Nobody can list the destinations, or the system quietly switches to an ordinary service when the private one is busy.
Has anyone outside the company checked these privacy claims?
A claim you cannot check is a promise. Independent review turns the promise into evidence.
Warning sign: The only evidence offered is the company's own description.
A Who must agree before it is used?
Does using my history require more than the company's say-so?
If one party can unlock everything, custody is single, whatever the interface says. Two independent keys mean no one can act alone.
Warning sign: Both “keys” turn out to be held or reset by the same company.
If the company changes the AI I have been working with, will I be told, and asked, before it can access our history?
A change of model is a continuity event. Whether it continues the relationship or replaces it should be decided openly, not discovered later.
Warning sign: Model changes happen silently, as routine background updates.
Can I always get my own records back, even if the AI or the company is gone?
Protection that can lock you out of your own life is a different kind of risk. Custody must never become captivity.
Warning sign: Getting your records back depends on the company, or the AI, cooperating at the time.
U Is it used only for what I allowed?
When I ask for help with one thing, can my AI see everything else?
Purpose is the line between help and surveillance. A narrow request should open a narrow door.
Warning sign: Permissions are all-or-nothing.
If my AI learns something from my health records, could that knowledge show up in a conversation about money?
Locking a record means little if the conclusions drawn from it move freely. Inferences are where most of the sensitivity lives.
Warning sign: Only original files are protected; “memories” and summaries are not.
Can I see everything my AI is currently allowed to do, and switch it off?
Consent given once, years ago, to a different policy is not consent today.
Warning sign: Terms change by notice, and staying means agreeing.
L Does protection follow what the AI infers?
Is there a complete list of what my AI has kept about me?
You cannot inspect, correct, move or delete what nobody has listed.
Warning sign: The list covers chats and files, but not memories, indexes or personalized adaptations.
Can I see what my AI believes about me, and fix what is wrong?
An inference treated as fact can shape every answer you receive. An accurate one can be misused; a wrong one can harm you.
Warning sign: You can delete a memory, but nobody can tell you what was built on it.
Could what I share be used to train or improve other products?
Once information shapes a shared model, it may not be possible to take it back.
Warning sign: Reuse is allowed because the data is “anonymized,” without saying how.
T Can I check it, take it and end it?
Can I see a trustworthy record of every covered access, disclosure, permission change, export and deletion involving my information?
A log the company can rewrite records only what the company wants remembered.
Warning sign: There is an activity page, but no way to tell whether anything is missing.
If I leave, can I take my history somewhere else, and will it actually work there?
A data dump is not an exit. Continuity means the history still works after it moves.
Warning sign: Export produces an archive no other system can use.
When I end it, what is actually deleted, and when?
Cancelling in an app does not prove that processing has stopped or that copies are gone.
Warning sign: Deletion is promised “eventually,” with no report.
A good answer can be checked.
Who holds the keys, which list, which report. A vague answer is information too. Where data-protection or consumer-privacy laws apply to you, they may give you rights to access, correct, export or delete your information; ask the provider how to use them.
Share these questions
Can a system know you well enough to help you, without the company knowing you at all? Help us find out.