Responsible Use of Generative AI at Work · Unit 4 of 5 · about 8 minutes

Unit 4: Disclosure, attribution and passing-off

The one idea this unit exists to teach

The question is never "did I use AI?", it is "would the person receiving this feel misled if they knew how it was made?". Honest use survives discovery. Passing-off does not. (We borrow "passing-off" from trade law, where it means selling goods dressed up as someone else's; here it means work dressed up as thinking that was not yours.)

Where AI assistance becomes a problem

Using AI to help produce work is now ordinary, like using a spellchecker. It becomes a problem in three situations:

Misleading the recipient. Work presented as your independent expert judgement or your original writing, when the substance came from a tool. A client paying for your professional opinion is buying your accountability, not a model's output with your signature.
Breaching an explicit rule. Academic coursework, examinations, some tender submissions, some publications and regulatory filings regulate AI assistance directly. Where a rule exists, the rule decides, not your judgement.
Republishing others' work. AI output can reproduce or closely paraphrase copyrighted material without flagging it. Publishing AI text verbatim under your brand carries plagiarism and copyright risk that you, not the tool provider, will own.

The disclosure test

A practical standard that works in almost every setting: disclose AI assistance whenever the recipient's trust in the work depends on how it was made.

Nobody needs to know a tool tidied your email grammar. A student submitting an AI-written essay, a consultant delivering an AI-generated report as bespoke analysis, and an author publishing AI text as their own writing are all on the wrong side of the same line: the recipient believed they were getting something they were not.

When in doubt, a single sentence usually resolves it: "Drafted with AI assistance and reviewed by [name]". If writing that sentence feels impossible in your context, that feeling is the answer; either the use is not appropriate there, or the rules require more than a sentence.

A note on your organisation's voice

Public statements and anything carrying your brand deserve one extra check: AI tools will happily produce generic claims your organisation cannot substantiate ("industry-leading", "trusted by thousands", "guaranteed results"). Passing-off is not only about authorship; it includes letting a tool put claims in your organisation's mouth that nobody signed off.

Scenario: the tender response

A consultancy submits a tender response with a methodology section drafted by AI and lightly edited. The tender rules require declaration of AI-generated content. A rival recognises the telltale generic structure and challenges the submission. The consultancy's problem is not that it used AI; competitors did too, and declared it. The undeclared use breached the tender rules, which is now an integrity issue affecting far more than one bid.

Key takeaways

Honest use survives discovery; passing-off does not.
Explicit rules (academic, tender, publication, regulatory) always override your own judgement about disclosure.
Never publish AI text verbatim under your name or brand without review.
The one-sentence disclosure resolves most doubt; the inability to write it is itself the answer.

Knowledge check

Q1. Which situation is acceptable WITHOUT any disclosure?

Q2. Your team wants to publish an AI-drafted thought leadership article under the company brand. What is the minimum responsible path?