Can I AI-Colorize Public Domain Photos in 2026?

What “Public Domain Photo Rules” Actually Mean

Yes, you can usually colorize a photograph whose underlying photograph has entered the public domain, and you can often publish the result commercially. The important distinction is that public-domain status attaches to a particular version of a work, not automatically to every photograph, adaptation, restoration, or later derivative created from it. A public-domain painting does not make every reproduction of that painting public domain, and a public-domain photograph does not necessarily make every digital scan, crop, caption, or platform copy freely reusable. The copyright status of the underlying artwork, the photograph itself, and your new colorization may therefore need separate analysis.

Also worth reading: How Do You Colorize Old Photos With AI Without Creating Fake Details? · Who Owns the Rights to AI-Colorize Photos, and Is It Legal? · How Should You Ethically Colorize Historical Black-and-White Photos With AI?

The date shown on an image is only a starting point. In the United States, works published by a qualifying organization with a notice are generally protected for 95 years from publication or 120 years from creation, whichever expires first. Different rules apply to unpublished works, government works, works created by employees in the course of employment, foreign works, and works restored from unpublished or unpublished source material. U.S. public-domain rules also change through legislation and court decisions, so a 1928 work should not be treated as public domain solely because 2026 minus 1928 is 98. For an international project, the law of the country where the photograph was first published, where your publisher operates, and where infringement is alleged may all matter.

A practical rule is to document the source, identify the exact image, and preserve evidence of its publication history. Colorization is generally safer than removing a watermark or copying a photographer’s current digital file, because the latter can involve rights in the photograph even when the subject is old. Public-domain status does not grant a right to mislead viewers, impersonate a living person, violate privacy, or infringe trademarks. For colorizethis.io users, the safest workflow is to use clearly documented public-domain source material, avoid copying recognizable copyrighted photographs, and obtain permission when the file provider asserts a claim over the particular reproduction.

Why a Public-Domain Painting Is Not Necessarily a Public-Domain Photograph

The familiar example of an old painting creates an avoidable trap. If an artist’s painting has expired, anyone may reproduce the painting, but a modern photographer may have created a new photograph of that painting. The photograph can be a separate copyrighted work because it selects lighting, framing, focus, perspective, color rendition, and other expressive elements. German courts have reportedly treated photographs of public-domain artworks as potentially protected, illustrating that the legal result can vary by jurisdiction. This is not a universal rule, and it should not be stated as though every court in every country reaches the same conclusion.

Museums and archives may also own rights in photographs they commission. A collection object, its catalog record, its 3D scan, and the museum’s website photograph may have different rights statuses. A museum may allow access to an object while retaining exclusive rights in its photography, database, or branding. Conversely, some institutional repositories make images available under open licenses, such as Creative Commons, even when the license grants only certain uses. An open license is different from a public-domain dedication, and an attribution requirement or non-commercial restriction can affect a commercial colorization project.

The reverse problem also occurs. A photograph can be public domain while a depicted person, building, logo, or later editorial annotation has separate rights. Copyright does not normally protect the mere subject of a photograph, but trademark law can restrict uses that confuse consumers about sponsorship or affiliation. Personality and privacy rights can matter in some jurisdictions, particularly where a recognizable person appears in a contemporary context. The copyright answer is therefore often straightforward while the broader publication decision remains context-dependent.

## How to Check a Photograph Before Coloring It

Start with the earliest reliable source for the photograph, not the website where you found a copy. Search the title, creator, collection number, date, and reverse-image information in institutional archives, museum databases, government collections, and rights statements. Look for wording such as “public domain,” “no known copyright restrictions,” “public domain dedication,” or a named open license. A vague statement that a website is “educational” or “copyright-free” is not enough. The strongest record usually connects the exact file to an author, publication date, source institution, and license or legal code.

For U.S.-first projects, check the publication history and whether the photograph was published with a copyright notice. The U.S. Copyright Office explains that photographs created after February 28, 2009, generally need a notice to receive the special statutory term for works published with notice, while many other jurisdictions protect photographs automatically from creation. Works created before that date can also have different notice and registration effects. If the image came from a foreign publication, consult the relevant rule in that country rather than assuming the U.S. calculation controls everywhere. Where a work is restored, renewed, or found in an archive after being “unpublished,” a special statutory provision may apply.

Next, inspect the actual file. A watermark or repeated stock-agency mark may signal that the image is licensed, even if the underlying scene is ancient. Do not remove it merely because the source is a public-domain artwork. Search the photographer’s or agency’s archive to see whether the file was released under a commercial license, editorial-only license, or a limited-use agreement. Stock photography is licensed rather than generally sold as ownership of copyright; an image from a stock agency may display only a small portion of a work and may have model-release or property-release restrictions.

After verification, save the source URL, access date, rights statement, and a screenshot or archived copy of the page. Record any changes you make, including crops, cleanup, facial adjustments, background additions, and color choices. This creates a defensible audit trail. If the status is uncertain, ask the archive, rights holder, or publisher for written permission before launching a paid product, app, or advertising campaign.

Can You Sell AI-Colorized Versions of Public-Domain Images?

Usually, yes, if the original photograph and all material you use are genuinely free of relevant restrictions. Selling a print, digital download, poster, mobile wallpaper, or commercial use in an AI-colorization service is generally different from reproducing the source photograph in an editorial archive. A colorized version can have commercial value because of the color choices, presentation, selection, editing, and service involved, provided that those additions do not copy protected material or breach a license.

The word “AI” does not eliminate copyright risk. Some tools have terms that grant permission only for personal use, limit commercial output, or prohibit using generated material as a standalone competing service. Other tools claim rights over output while saying that ownership will follow local law, leaving the actual result uncertain in jurisdictions where purely machine-generated works may not receive copyright protection. Training data can create disputes too, but the fact that an image is publicly available online does not automatically make it legally usable in every dataset. The source provider’s terms, the model’s terms, and the law of the relevant jurisdiction all need attention.

Do not imply that a business owns exclusive copyright in every colored version of a historic image unless your jurisdiction recognizes the necessary human authorship and the work meets applicable originality requirements. Use truthful language such as “AI-colorized interpretation based on a public-domain photograph,” especially when a viewer could otherwise think the result is an authentic historical color record. For living subjects, avoid creating a colorized portrait that falsely appears to document a real event, especially where the depiction could affect identity, reputation, or political persuasion.

Comparison of Public-Domain, Open-License, and Permission-Based Use

Public-domain use is usually the cleanest route, but it is not the only route. An open license may permit commercial use if its conditions are met. Creative Commons licenses can include attribution, share-alike, non-commercial, and no-derivatives terms; a derivative image may also need to be distributed under the same license. A public-domain mark is generally simpler because it does not impose those conditions, though other rights can still apply. Permission-based licensing is often more expensive, but it may be preferable when provenance is uncertain or a recognizable contemporary subject creates reputational risk.

Cost is often the deciding factor for small operators. Public-domain and appropriately licensed source images may cost $0, while archive fees, scan charges, stock licenses, legal review, and commercial image subscriptions can range from a few dollars to hundreds or thousands of dollars per project. A tool subscription may add another monthly or annual fee, but the tool price does not clear the source rights. Before publishing, budget for attribution, quality control, records, and customer support rather than treating the raw image as the only expense.

Common Mistakes and Legal Pitfalls

The most common mistake is confusing age with status. A photograph can be old but still protected because it was first published recently, contains a separate scan or restoration, or falls under a special restoration rule. Another common mistake is trusting a social-media repost that says “free to use” without explaining whether the claim covers the image file, only the subject, or only personal use. Search results can also mix the original work with derivative photographs, cropped versions, and AI recreations that are not the source you intended.

Second, users sometimes assume that public-domain photographs cannot be copyrighted in a new version. A colorization can be an adaptation depending on how substantially it changes the original and the governing law. The fact that an image is a new file does not automatically make it independent, especially where the same composition, pose, lighting, and framing remain identifiable. The safer approach is to use a genuinely different composition or obtain advice when the plan depends on a particular country’s adaptation rules.

Third, do not remove attribution, watermarks, or rights-management information. Even if a particular use is legally possible, deleting a notice can violate contract terms, open-license conditions, or platform rules. Fourth, avoid presenting AI-generated color as verified historical evidence. In the Ansel Adams example, “Moonrise, Hernandez” was a 1942 black-and-white photograph, while a later colorized version attributed to AI created a dispute because viewers and institutions could confuse the reinterpretation with an authentic Adams work. The controversy demonstrates that labeling matters even when copyright questions are separate from authenticity and public trust.

Finally, privacy and publicity rules are not replaced by copyright. Historical subject matter can still create ethical or legal concerns, and a modern person’s likeness may be protected in some places. Commercial use of a public-domain image is not a blanket defense against defamation, fraud, or misleading advertising. Clear captions, disclosures, and moderation policies are inexpensive controls.

When to Act, and What to Document Before Launching

For personal experimentation, colorize freely only when the source and tool terms appear compatible with the intended use. For a public website, archive the source record and use a visible credit line. For a commercial storefront, do the work before ordering inventory or advertising: verify the source, check commercial permissions, inspect the tool’s terms, and save evidence of the human decisions behind the result. The longer the project takes, the more important a dated rights record becomes.

A sensible threshold is to stop and seek permission if the source lacks a clear license, the image comes from a commercial stock library, the photograph is a modern photograph of an older artwork, the subject is identifiable and contemporary, or the output will be sold in multiple countries. Another trigger is an institutional watermark or statement reserving rights in photographs, even if the underlying artwork is public domain. Written permission should identify the exact image, permitted edits, commercial channels, territory, duration, attribution wording, and whether the permission extends to sublicensing.

Keep a rights log containing the source creator, title, date, repository, original URL, license text, access date, file hash where practical, edits made, software or model used, and the date the output was generated. If the service uses third-party vendors, confirm whether they claim rights in prompts or outputs and whether data is retained. On privacy grounds, avoid uploading identifiable people’s images to a consumer tool unless the service’s data policy has been reviewed. These steps do not guarantee a favorable legal result, but they reduce the chance of an avoidable mistake and make a takedown response much easier.

A Defensive Workflow for colorizethis.io

The best site policy distinguishes verified public-domain sources from merely online images. It can offer users a choice between an informational archive, personal-use experimentation, and a commercial project, while refusing to guarantee clearance from a thumbnail or search result. For every accepted project, the site should record the source and label whether the status is verified, licensed, uncertain, or permission-based. A user-facing notice can state that a public-domain underlying work does not automatically clear the digital photograph or all related rights.

The site should also explain that AI colorization is interpretive. It can improve visual access and help people engage with old photographs, but it can introduce historically inaccurate skin tones, clothing colors, weather, lighting, or architectural details. For educational material, offer an “AI interpretation” label and distinguish it from a color record created by the photographer or a documented restoration. This is particularly important for archival, museum, historical, and public-history contexts.

A practical release process would use four gates: source identification, license review, human quality and authenticity review, and publication labeling. If any gate fails, the upload can be held for permission or rejected. Users should not be promised that an image is “copyright-free” merely because its creator died more than 70, 95, or 100 years ago. Instead, state the jurisdiction, term calculation, source evidence, and remaining limitations. That wording is less dramatic than a blanket promise, but more dependable for a real AI-colorization product operating in 2026.