Privacy, Consent, Copyright, and Honest Attribution

Know what material you have, what you may do with it, and how to describe the result.

A volunteer sends you a photograph for a newsletter. You want to brighten it, animate the people in it, and add a familiar voice reading the story.

At first, this seems like one creative task. It actually involves several questions. Who controls the photograph? What use was permitted? Did the people agree to this new portrayal? Where will the voice come from? Which service will receive the files?

Possessing a file does not answer all of those questions. A good workflow keeps the source, intended use, permissions, and disclosure attached to the material as it moves through the project.

Begin with an asset map

An asset is any piece of material you use: text, an image, a voice recording, a spreadsheet, a logo, or a generated result.

For each asset, identify its source and what you intend to do. Reading a document internally, uploading it to a service, adapting it, and publishing it are different uses.

Also identify people and private information. A photo may show a person who did not create the image. A recording may contain someone else's voice. A spreadsheet may contain contact details that are irrelevant to the writing task.

Ask an assistant to help make an inventory using descriptions of the material where possible. You do not need to upload the sensitive file merely to ask which permission questions it raises.

Separate the permissions

Imagine the following fictional project: a community organization wants an article, a header image, and an audio version for its public website.

These are the facts of a fictional exercise. They illustrate how to compare a proposed use with the permission actually supplied. A real case may involve additional contracts, exceptions, institutional rules, or local law.

Do not fill an empty permission field with “probably fine.” Mark the gap and choose a concrete way to resolve it.

A license answers some questions, not every question

Creative Commons Attribution 4.0 permits sharing and adaptation, including commercial uses, subject to its conditions. Those include appropriate credit, a license link, and indicating changes. Its deed also warns that privacy, publicity, or moral rights may affect an intended use. Creative Commons: CC BY 4.0.

That is why the chart row still requires inspection. Is the license actually attached to this chart? Does it cover all components? Are there separate credited elements? What attribution is supplied?

Do not assume that every image found through a search carries the same license as the page around it. Record the source and terms for the particular asset.

Likewise, credit alone does not establish permission for every use. Acknowledging a photographer is not a substitute for checking the rights involved in altering or publishing the photograph.

Understand the U.S. copyright reference carefully

The U.S. Copyright Office's January 2025 report explains that protection for work involving generative AI depends on sufficient human authorship of expressive elements. Human selection, arrangement, or modification can matter; merely supplying prompts does not by itself establish authorship of the generated output. AI assistance does not automatically disqualify the human-authored work. U.S. Copyright Office: Copyrightability report announcement.

This is a U.S. reference, not a universal rule for every jurisdiction or every asset. It also addresses a different question from whether your input material was lawfully used or whether a depicted person agreed to a portrayal.

Do not turn “I edited it” into a blanket guarantee of exclusive rights. Preserve what the human actually contributed and examine the intended use. For a consequential unresolved rights question, obtain qualified review of the specific material and jurisdiction.

The Office maintains the report series and related guidance on its Copyright and Artificial Intelligence page. Check current guidance when preparing a real registration or publication decision.

Give the tool only the information the task needs

Suppose the article thanks volunteers as a group. It does not need their home addresses, personal phone numbers, or private scheduling notes.

Prepare a limited input: the approved event facts, the intended audience, and any quotations cleared for the proposed use. Keep confidential records in their appropriate system.

Permission to work with information inside an organization does not automatically settle whether it may be uploaded to an outside service. Check the organization's approved tools, data-handling rules, and the service terms that apply to the account.

Use placeholders or fictional examples when exploring a workflow. When actual personal information is necessary, establish the appropriate basis and handling before processing it.

Preserve the human contribution and the source trail

Keep useful drafts, source notes, and significant edits. Record which facts you supplied, which passages you wrote, and what the assistant proposed.

This record helps you review accuracy, explain the process, and identify material that still requires attention. It does not create rights that were absent or prove that every claim is correct.

For the fictional article package, a simple provenance note might identify the original author, approved event fact sheet, licensed image source, final editor, and narration method.

If a generated illustration resembles a documentary photograph, label its role clearly. “Illustration of a possible future classroom” tells readers something different from “Students at our November workshop.” Do not give a synthetic scene a factual caption implying it records an event that occurred.

Disclose the process in useful language

A disclosure should describe what happened. “AI-assisted” may be too vague when readers need to know whether a voice, image, or passage was generated.

For a hypothetical article where the human researched, directed, checked, and revised an AI-assisted draft, a truthful note could read:

The author used AI to organize source notes and develop an initial draft, then checked the cited evidence and revised the final text.

Use that sentence only if it accurately describes the work. If source checking remains incomplete, do not claim it happened. If a synthetic voice was used, describe that separately where relevant.

Disclosures may be required by a publisher, institution, platform, contract, or applicable law. Even where a particular note is not required, it can help readers distinguish an illustration, simulation, or reconstruction from a record of real events.

Students: permission, citation, and authorship are different questions

A student may have permission to use AI for feedback while being required to write the final argument independently. A citation to the tool does not make an otherwise prohibited use acceptable.

Check the assignment and institutional rules. Keep source citations for evidence separate from a process note about AI assistance. Describe the actual help received: brainstorming, feedback, translation, coding assistance, or drafting.

For group work, agree on permitted use and avoid uploading classmates' contributions or personal information without the appropriate basis. For research participants, follow the project's consent and data-handling requirements rather than substituting a generic prompt.

Cornell's guidance emphasizes communicating expectations and verifying AI-assisted work when use is allowed. Your own institution's rules govern your assignment. Cornell: AI & Academic Integrity.

Your exercise

Use the fictional asset table to prepare a publication-ready alternative plan. Keep the original writing workflow, replace or clear the image use, choose authorized narration, supply the chart's required attribution, and omit the private appendix.

For each asset, record source, intended use, permission or other applicable basis, attribution, unresolved issues, and final decision. Draft a process disclosure that makes no claim about work that has not happened.

Completion check: Every included asset has a documented source and intended-use basis, unresolved questions remain visible, and the process description matches reality.