Free tools Windows power users keep installed
One-click scans. No signup required.
ThatPainter is reader-supported. When you buy through links on our site, we may earn an affiliate commission. Learn More
The COPIED Act is not law. As of August 18, 2026, the current version—S. 1396—is an introduced Senate bill, not an enforceable federal requirement. It would not make removing every digital watermark illegal. Instead, it proposes rules around machine-readable content provenance, deceptive tampering, platform preservation, AI-generated content and certain commercial uses of provenance-linked material for AI training.
The bill was first introduced as S. 4674 in 2024 and reintroduced as S. 1396 on April 9, 2025. Neither version has passed Congress or been signed into law.
What does COPIED stand for?
COPIED stands for Content Origin Protection and Integrity from Edited and Deepfaked Media. The proposal addresses how the origin and editing history of digital content— including paintings photographed for publication, illustrations, photographs, text, audio and video—could be recorded and preserved.
The Tool Desk
Outbyte PC Repair FREEClear out junk files and repair common Windows errorsFree Scan →Outbyte Driver Updater FREEScan for outdated or missing drivers - takes under a minuteDriver Scan →#1 Best Overall
- Used Book in Good Condition
Its policy goals include improving attribution, making synthetic or modified media easier to identify, discouraging deceptive manipulation of provenance records and giving creators more control over some commercial AI uses of their work.
Which COPIED Act is being discussed?
The original bill, S. 4674, was introduced on July 11, 2024, during the 118th Congress. A substantially similar successor, S. 1396, was introduced on April 9, 2025, during the 119th Congress.
Current coverage should refer primarily to S. 1396. The 2024 bill is not an active 2026 bill simply because it had the same name; bills generally must be reintroduced in a new Congress. The current legislative record lists S. 1396 as Introduced and referred to the Senate Committee on Commerce, Science, and Transportation. It has not been enacted.
What would the bill actually regulate?
The central concept is content provenance information: state-of-the-art, machine-readable information documenting the origin and history of digital content. That could include information about who created an asset, what tool produced it, whether it was edited and what changes were made.
The bill is technology-neutral in its wording. It does not simply define provenance as a logo placed visibly over an image, and it does not name C2PA as the only permitted system.
Would removing a digital watermark become illegal?
Not as a blanket rule. The bill would target knowing removal, alteration, tampering with or disabling of content provenance information when the conduct is in furtherance of an unfair or deceptive act or practice in or affecting commerce.
That is materially narrower than the claim that “watermark removal is illegal.” The legal analysis would depend on the type of information, the actor, what the person knew, the commercial context and whether the conduct was connected to deception.
For example, deliberately stripping provenance from an artwork before falsely presenting it as an original creation would raise a different issue from a routine image conversion that unintentionally discards metadata. The bill text does not resolve every technical failure involving uploads, screenshots, compression, cropping, exports or reposts.
Quick wins for a faster PC:
Clear out junk files and repair common Windows errorsFree Scan →Fix the driver behind crashes, sound loss and screen glitchesFind Drivers →Covered platforms would also generally be restricted from removing, altering, tampering with or separating provenance information from content so that users can no longer access it. The bill includes an exception for action that is necessary, proportionate and limited to security research.
Watermarks and provenance are different technologies
| Technology | What it does | Typical limitation |
|---|---|---|
| Visible watermark | Displays a human-readable name, logo or ownership notice. | Can be cropped, covered or obscured. |
| Metadata | Stores technical or descriptive information in or alongside a file, such as EXIF, XMP or IPTC data. | May be discarded during conversion, export or upload. |
| C2PA provenance | Uses signed, machine-readable records to describe origin and editing history. | Can be lost, invalidated or misunderstood as proof of truth or ownership. |
| Invisible pixel watermark | Embeds a signal in image or video data. | May degrade under resizing, compression or other transformations. |
| Statistical AI watermark | Uses detectable patterns in generated text, images or other output. | Detection and robustness can vary by content and transformation. |
A provenance record may provide evidence about origin, signing or editing history. It is not automatically a copyright registration, conclusive proof of ownership or proof that a particular use infringes copyright.
What would it require from platforms?
The bill defines a covered platform as a website, internet application or mobile application available to users in the United States that either generates at least $50 million in annual revenue or has at least 25 million monthly active users for at least three of the preceding 12 months.
For platforms meeting that definition, the proposal would generally require preservation and accessibility of attached or associated provenance information. That raises practical questions that the bill does not completely answer:
Do these 3 things before closing this tab:
1Repair Windows errors before they cause bigger problems2Fix the driver behind crashes, sound loss and screen glitches3Clear out junk files and repair common Windows errorsRank #3
- Whether ordinary transcoding that accidentally destroys a manifest creates liability.
- How a platform should handle screenshots, screen recordings and derivative exports.
- When moderation, privacy protection or anti-malware processing justifies changing or removing data.
- What happens when a re-upload arrives without its original provenance record.
Technical standards, later regulations, litigation and the final language of any enacted version would determine how those situations are treated.
What would AI tool providers have to do?
Beginning two years after enactment, a person commercially making available an interstate-commerce tool primarily used to create or substantially modify covered content would have to give users the ability to include content provenance information.
If a user chose to include that information, the provider would also have to use reasonable security measures—where technically feasible—to make it machine-readable and not easily removed, altered or separated from the content.
This is an important distinction: the current proposal does not plainly require every AI output to carry a universal, indelible watermark. It would require qualifying tools to offer a provenance capability, with additional security duties when the user opts in.
The Tool Desk
Outbyte PC Repair FREERepair Windows errors before they cause bigger problemsFix Now →Outbyte Driver Updater FREEScan for outdated or missing drivers - takes under a minuteDriver Scan →How could the bill affect AI training?
The proposed training restriction is linked to provenance and commercial conduct. It would prohibit a person, for a commercial purpose, from knowingly using covered content carrying provenance information to train an AI or algorithmic system or generate synthetic or synthetically modified content without:
- Express, informed consent from the content owner; and
- Compliance with the owner’s applicable terms of use, including compensation terms where required.
The provision also addresses content from which the person knows, or should know, provenance information was removed or separated in violation of the bill.
This would not be a universal ban on training AI with every copyrighted work. The proposed rule is tied to several conditions: the content must be covered, the use must be commercial, the conduct must involve the required knowledge, provenance must be relevant and consent and terms of use must not have been satisfied.
That means the proposal would not by itself settle every dispute over fair use, copyright ownership, licensing, contracts or the use of publicly accessible material in datasets. The U.S. Copyright Office’s AI initiative separately examines copyrightability, training, digital replicas and related legislative questions.
Is C2PA required?
No current text establishes C2PA as a legally mandatory implementation. C2PA is an open standard for content provenance and is a likely technical reference point for organizations building these systems, but the bill describes provenance information and consensus-based standards in technology-neutral terms.
The Senate Commerce Committee’s 2024 summary describes an approach involving voluntary, consensus-based standards for watermarking, provenance, synthetic-content detection, AI-training data practices, evaluation and cybersecurity.
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.How is COPIED different from an AI watermark mandate?
A separate proposal, the Advisory for AI-Generated Content Act, was introduced as S. 2765 in 2023. That bill would have required AI-generated materials to include a watermark meeting standards established by the Federal Trade Commission. It also remained at the introduced stage.
The COPIED Act is broader in subject matter. It focuses on origin and editing history, provenance preservation by large platforms, user-enabled provenance in qualifying creation tools and certain provenance-linked commercial AI uses. It should not be reduced to a requirement that every AI image, painting or text output contain one visible watermark.
Free tools Windows power users keep installed
One-click scans. No signup required.
What does this mean for artists and painters?
For artists, photographers and publishers, provenance systems could be useful as documentation and attribution tools even if they cannot prevent copying. A signed record may help show how an image was created or edited, while a visible watermark can still serve the separate purpose of making attribution obvious to human viewers.
However, provenance is not an anti-copying barrier. A screenshot, flattened export or photograph of a canvas may no longer carry the original machine-readable record. Provenance can also create privacy concerns by exposing a creator’s identity, location, software, timestamps or editing history. Anonymous or safety-sensitive creators may not want every piece of content tied to their real identity.
Creators considering a provenance workflow should distinguish among three goals:
- Attribution: making the creator’s name or studio visible.
- Documentation: preserving information about origin and edits.
- Enforcement: building evidence for a dispute or rights-management process.
No single watermark or provenance system guarantees all three.
What could change if the bill passes?
- Congress would first need to pass the bill in both chambers.
- The president would need to sign it, or Congress would need to enact it through another constitutional process.
- Implementation would follow the statute’s timing. The tool-provider requirements described in the current text would begin two years after enactment.
- Standards and technical practices would develop. Platforms and tool providers would need to determine how to preserve, expose and secure provenance records.
- Courts and regulators would clarify edge cases. Questions about knowledge, deception, technical feasibility, privacy, contracts and AI training would likely require interpretation.
Until those steps occur, the bill does not impose new federal obligations on artists, platforms or AI companies.
Potential benefits and concerns
Why supporters may favor the approach
- More reliable attribution and auditability.
- Greater transparency about editing and synthetic generation.
- More leverage for creators negotiating commercial AI-training permissions.
- Less opportunity to strip provenance and misrepresent content.
Why implementation may be difficult
- Provenance is not the same as truth; a valid signature can record what a signer claimed without proving the claim is accurate.
- Ordinary file handling can destroy metadata without deliberate tampering.
- Privacy-sensitive creators may not want identifying information embedded in their work.
- Small developers may face uncertainty and compliance costs.
- Multiple provenance layers may disagree, including metadata, C2PA records, visible labels and statistical watermarks.
- AI-training disputes could overlap with copyright, contract law, fair use and constitutional challenges.
Practical takeaway
For now, artists should not assume that removing a watermark is a federal crime, and AI developers should not assume that the COPIED Act already bans training on online artwork. The proposal is better understood as a provenance-linked framework: it would address knowing deceptive tampering, require qualifying platforms to preserve provenance, give users of certain tools a way to include it and restrict some commercial AI uses of provenance-marked content without informed consent and compliance with terms of use.
Organizations evaluating provenance tools can review the C2PA standard, Adobe Content Credentials, Truepic and Digimarc. These services differ in purpose and implementation, and none can guarantee that provenance survives every screenshot, export, transformation or platform upload.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
