12.1.1Their data is your responsibility
It is tempting to think of user data as an asset, a pile of value you are accumulating. Flip that. The moment it is real people's data, it is a liability: something you now have to protect, keep accurate, and be ready to hand back or delete.
That changes who a mistake hurts. A leak is not your inconvenience, it is a real person's home address in the open or their reused password in an attacker's hands. You are holding it on their behalf, so their exposure is your responsibility.
12.1.2Collect only what you need
Every field you collect is one more thing to secure, keep correct, and answer for if it leaks. So collect the least you can. This is data minimization, and it is the cheapest security control you have: data you never stored cannot be stolen.
Be strictest with personally identifiable information (PII), the data that points at a specific human, like their email, phone, or home address. Ask of each field whether the app truly needs it now, or whether you kept it just because it was easy.
Rule of thumb: If you cannot name why you store a field, do not collect it.
12.1.3Know what you hold and why
You cannot protect what you have not written down. A data inventory is one plain list of every kind of data you store, why you keep it, and how sensitive it is. You already sketched the things your app stores when you modeled your data; the inventory adds the why and the risk.
Have your agent produce it and keep it in your system. It looks like this:
That one page tells you where to spend your protection effort, and it is the first thing you reach for when something goes wrong.
12.1.4Clean data pays you back twice
Keeping this data lean and labeled is not only about risk. The same clean, well-kept data is what your AI reads to make decisions and trace bugs once the app is live. That is the second payoff the Operate part builds on, when you are watching the app run. Kept clean now, it pays twice.
This prompt has your agent draft the inventory from your app:
Do this now: paste the prompt, let your agent draft your data inventory, and delete any field it cannot give a clear reason for.
