AI Guides › Judgement & Guardrails
By Nigel Guy · 2 min read
"The AI drafted most of it" gets said as though it explains something about who's accountable if the finished piece turns out to be wrong, unfair, plagiarised, or badly reasoned. It doesn't. The moment you submit it, send it, publish it, or act on it, the share of it that came from a model stops being a meaningful fraction of anything — responsibility for the output doesn't dilute in proportion to how much of the drafting you didn't do yourself.
The rule: your responsibility for anything you submit, send, or act on is total the moment you do it, regardless of what proportion of it an AI tool produced — there is no partial-credit version of accountability.
Before anything AI-assisted goes out under your name, answer these honestly:
| Question | What a "no" means |
|---|---|
| Did you personally verify the factual claims in it? | If not, you're vouching for something you haven't checked. |
| Do you actually understand what you're submitting, well enough to explain it unaided? | If not, you can't defend it if it's questioned. |
| Would you defend this exact content if someone challenged a specific line in it? | If not, that line needs work before it goes out, not after. |
| If this turned out to be wrong, would "AI helped write it" change how you'd feel about having sent it? | If it would, that's worth noticing before you send it, not after. |
Skip the instinct to either over-credit or over-blame the tool. Claiming work as entirely your own original thinking when a model did substantial drafting is its own kind of dishonesty in contexts where that matters — academic work, attributed writing, anywhere originality itself is the claim. But citing the tool as a shield the moment something in the output turns out to be wrong is the opposite failure, and neither one is what disclosure is actually for.