A tool, not a substitute for responsibility
Generated media is a catch-all for images, audio, video and text produced by software in response to instructions. Used well, it can speed up drafts, spark ideas, help with accessibility tasks and make some projects possible that would otherwise be out of reach.
Used carelessly, it can spread errors, misrepresent people or blur who is accountable for the work. The principle that holds across cases is this: whatever you publish is yours. The audience will hold you responsible, not the tool.
Specific products and features change quickly, so this article sticks to principles that should outlast them.
Principle 1: Be honest about what it is
Do not present generated work as something it is not. A generated image is not a photograph of an event. A generated voice is not a recording of a person. A machine-written draft is not firsthand experience.
Telling people is the practical form of this honesty, and AI disclosure covers wording and placement in detail. Also avoid passing off generated work in settings where it is prohibited, such as certain classes, competitions, publications or client contracts. Read the rules before you begin.
Principle 2: Verify before you trust
Generated text can sound fluent while being wrong. It may invent details, quotes, sources or events, and present them with confidence. Images may contain impossible details, and summaries may drop important qualifications.
Build checking into the process:
- Treat factual claims as unverified until you confirm them against reliable sources, using your own self fact-checking routine.
- Never publish a name, date, number or quotation you have not verified.
- Be especially careful with health, money, law and safety, where errors can harm people. Present general information, avoid promises and point readers to qualified professionals or official guidance.
- Read the whole output. Mistakes tend to hide in the parts nobody examines.
Principle 3: Care for the people involved
Generated material can portray people, imitate them or reflect assumptions about them.
- Do not imitate real individuals without clear permission. Personal likeness and voice are especially sensitive.
- Do not create fabricated scenes of real people or present invented events as real. See the concerns around synthetic media.
- Review for stereotypes and gaps. The causes and checks are covered under output bias.
- Protect private information. Avoid pasting other people’s personal details, unpublished work or confidential material into tools whose handling of that data you do not understand. Read a tool’s terms and settings first.
Principle 4: Understand that rights questions are unsettled
Copyright and licensing for generated material is an area where law and practice vary by country and continue to change. Questions include whether generated output can be protected, whether it may resemble existing works, and how the terms of a particular tool treat ownership and commercial use.
Without giving legal advice, some sensible habits are:
- Read the terms of any tool you use and keep a record of them as they stood when you made the work.
- Avoid prompts that ask for the style or exact work of a named living creator in ways that would copy their expression. Consider the effect on the people whose work you are echoing.
- Check for resemblance. If output looks very close to something existing or contains recognizable logos or characters, do not use it as is.
- Do not assume you hold exclusive rights. If originality or ownership matters, such as for a logo or a client deliverable, use official guidance on copyright basics and seek qualified advice.
- Credit what you can. Where you used licensed assets such as stock media or material under open licenses alongside generated elements, keep proper attribution records.
Principle 5: Keep humans in charge of judgment
A useful working model treats output as raw material.
- Define the purpose first. A clear brief, whether a short note or a formal creative brief, helps you judge the result.
- Generate options, then choose. Do not accept the first result by default.
- Edit with your own voice and standards. Revision process steps such as structural editing and proofreading still apply.
- Check against your values. Ask whether you would be comfortable explaining how this was made.
- Own the decision. If something goes wrong, fix it openly, as covered under public corrections.
Reliance on automated drafting can also flatten your style or make work generic. Using tools for the parts that bore you, and keeping the thinking and the final judgment for yourself, tends to produce better work.
Principle 6: Respect fellow creators
Many tools learned from other creators’ work, and opinions and legal views on that differ. You can respect the debate by crediting and hiring human artists, writers and performers where it matters, and by being clear about what you made with and without tools.
Principle 7: Know when not to use it
Some uses are poor ideas whatever the capability of the tool: fake testimonials or reviews, fabricated evidence, impersonation, content aimed at deceiving or pressuring people, and material depicting real people in degrading situations. If a use only works because the audience will be fooled, that is a signal to stop.
A brief checklist
- Have I verified every factual claim?
- Does anyone appear who has not agreed?
- Is there anything that copies or closely resembles existing work?
- Have I read the tool’s terms for my intended use?
- Will the audience know what is generated where it matters?
- Would I stand behind this publicly?
Technology and rules in this area will keep shifting, and no checklist guarantees a perfect outcome. Honesty, verification, care for people and personal accountability remain sound anchors whatever the next tool looks like.