Authorship, IP & Disclosure

Start with the question that actually troubles people, because it is the one you can answer for yourself with confidence: is AI-assisted work still *your* work? The answer, for how you practise and how you should think of yourself, is yes - provided you did the authoring, which is a real thing and not a formality. Authorship in design has never been about who performed the mechanical act of production. Draughtspeople, junior architects, visualisers, model-makers and specialist consultants have always contributed to work that goes out under a principal's name, and the principal is rightly the author because they supplied the intent, the judgement, the selection and the accountability. Claude is another contributor of that kind - an extraordinarily fast and well-read one with no claim of its own.

What makes you the author is the shape of your involvement. You framed the problem. You gave the context only you had. You rejected what was generic, wrong or off-brief. You combined, corrected and pushed until the result expressed *your* thinking, not a plausible average of everyone's. And you put your name and your liability on it. Strip those away - accept a first draft unread, unedited, unjudged - and you have not authored anything; you have forwarded output, which is precisely the failure mode this whole course is built to prevent.

This is why the discipline of the earlier modules is also the guarantor of your authorship. The direct-draft-judge-refine loop is not only how you get good work out of Claude; it is what makes the work yours. The more of your judgement flows through it, the more genuinely it is authored by you. Used lazily, Claude produces un-owned, generic work that anyone with the same prompt could have made - and that, not the tool itself, is what hollows out authorship. Used as an instrument of your own thinking, it leaves the authorship exactly where it belongs: with the designer who decided.

Now the genuinely legal part, handled with appropriate humility. Copyright and intellectual property law around AI-assisted work is unsettled, varies significantly between countries, and is being actively litigated and legislated as of 2026. This lesson will not give you rulings, because anyone who gives you a confident, universal ruling on this is overreaching - and because a wrong ruling here is exactly the confident-plausible error the last lesson warned about. What you can usefully hold are the contours of the questions and the practical protections, and then take anything that matters to a qualified IP lawyer in your jurisdiction.

The contours look roughly like this. Copyright has traditionally protected works of *human* authorship, which is why the degree and nature of human creative input into an AI-assisted work matters to whether, and how, it is protected - and different jurisdictions are drawing that line differently and revising it. Purely machine-generated output with negligible human authorship sits in the most uncertain territory; work substantially shaped, selected and arranged by a human sits on firmer ground, but the details differ by country. Separately from copyright, there are the *terms* under which you used the tool: as of 2026 Anthropic's consumer terms address the rights in what you generate, and those terms are the thing to actually read for your account rather than assume. And there is the question of *inputs* - you should not feed Claude a competitor's copyrighted material or a client's confidential IP and treat the output as clean.

The practical posture that keeps you safe without a law degree: keep human authorship real and documented (your loop does this); read the current tool terms for your plan; do not launder others' IP through the model; and for anything commercially important - a design you need to protect, a dispute, a contract clause about IP ownership - get it in front of a lawyer. That is not evasion; it is the same principle as the last lesson. Claude, and this course, can explain and point; only the professional - here, counsel - can rule.

Disclosure - telling clients and collaborators how the work was made - is less a legal question than a professional-honesty one, and here you can act with confidence because it rests on principles you already hold. The guiding idea is simple: do not misrepresent how the work was produced, and do not claim a certainty your process does not support.

In practice, most everyday AI assistance needs no ceremony. Using Claude to tighten your prose, structure a report, summarise a byelaw you then verified, or draft a schedule you checked is no more disclosable than using a spellchecker, a specification library or a junior - it is ordinary professional tooling, and clients neither need nor want a footnote on every email. What does call for candour is anything where the *client's expectation* would be materially wrong without it: if a client believes they are commissioning wholly bespoke, hand-authored creative work and would feel misled to learn AI drafted the core of it, honesty asks that you not actively create that false impression. And if a client asks you directly whether and how you use AI, you answer truthfully - always.

The figure sets out a workable ladder from routine to explicit. The test running through it is the mirror test: would you be comfortable if the client watched exactly how this was made? If yes, carry on. If a disclosure would change how they feel, that discomfort is the signal to be open rather than to conceal. Two contexts deserve extra care: competitions and awards, which often have their own rules on AI use that you must read and follow; and contracts, where an explicit clause about AI use, IP ownership and confidentiality is increasingly wise and, again, a matter for your lawyer to word. Handled with ordinary honesty, disclosure is not a threat to your standing - it protects it. Clients trust designers who are straight with them about their methods, and that trust, once you have it, is worth more than any single project. Concealment, by contrast, is fragile: it only has to be discovered once to poison a relationship you spent years building, and in the age of AI it is increasingly likely to be discovered. Honesty is simply the cheaper, stronger and more professional choice.

These three threads are really one professional stance, and it is the same stance the whole course has argued for: stay genuinely the author, respect the rights around your work, and be honest about your methods. Held together, they turn what looks like a minefield into a straightforward way of working.

Stay the author by keeping your judgement in the loop - which you were going to do anyway, because it is also how you get good work. Respect IP by keeping human authorship real, reading your tool's current terms, refusing to launder others' protected material, and routing the commercially serious questions to counsel. Be honest by not misrepresenting how work was made, disclosing where a client's expectation would otherwise be materially wrong, answering directly when asked, and following the specific rules of competitions and contracts.

Notice what this stance protects: not the tool, but *you* and your relationship with your client. A designer who can say "this is my work; I directed and judged every part of it; I used AI as a drafting instrument and I checked what mattered; and I am happy to tell you exactly how" is in an unimpeachable position - creatively, legally and ethically. That is a far stronger place to stand than either of the two poles people drift to: the hype-merchant who lets AI produce un-owned work and passes it off as bespoke, or the frightened practitioner who hides their AI use as if it were shameful. Neither the concealment nor the pretence is necessary. Do the authoring, respect the law, tell the truth - and AI-assisted practice sits comfortably inside the professional integrity you already have. The next and final lesson takes this integrity and asks how to build a whole practice on it.