# Responsible AI

By [DYLIT Chronicles](https://dylit.info/user/dylitmediabuzz)

[Everything AI - beyond the hype](https://dylit.info/pr/everything-ai-beyond-the-hype/6a9efac02e92664f4d50cf9d) > [Responsible AI](https://dylit.info/ch/responsible-ai/6ab3685f99638e119f592dae)

The Do and Don't Card One page. Print it, pin it up, and you have covered most of what matters.   DO Anonymise before you paste anything in. Placeholders for names and figures. Put the real details back yourself, afterwards. Verify every specific detail. Names, numbers, dates, citations, quotes. The reasoning usually holds. The details are where it breaks. Take ownership of the output as your own work. If your name is on it, you wrote it. Whatever produced the first draft. Keep people involved in decisions that affect people. Hiring, evaluations, lending, discipline. A human reviews the outcome, every time. Ask when you are not sure. Before you paste, not after. Asking early is a normal question. Asking late is an incident report.   DON'T Never paste sensitive information into an unapproved tool. Customer data, employee records, credentials, unreleased financials. Never send output you have not read. Every word, start to finish. Skimming is how a fabricated figure reaches a client. Never cite a source you have not opened. Confirm it exists. Then confirm it says what the output claims. Two separate steps.   How to use this Keep it visible for a month. That is usually long enough for the habits to stop requiring thought. Each line is short because it is a reminder, not an argument. The reasoning behind every one of them is in the longer posts in this channel, and any line that seems arbitrary is worth reading up on rather than following blindly. Two of them tend to get quietly abandoned first: reading every word, and checking citations. Both get dropped under time pressure, which is exactly when errors are most likely to be in the document. If you can only hold onto two, hold onto those.