Generative artificial intelligence is reshaping how companies develop copy, logos, designs and invention disclosures. But the most important legal question is not privacy in the abstract. It is whether and to what extent an IP owner must disclose that AI helped create the asset.
Current U.S. law offers no single confidentiality rule for AI-assisted creation. Instead, each branch of intellectual property law draws its own line between compelled disclosure and process secrecy, and that line reflects the doctrine at the center of each regime: human authorship in copyright, human inventorship in patent law and source identification in trademark law.
Why the Disclosure Rules Diverge
The constitutional text itself helps explain the split. Article I, Section 8, Clause 8 empowers Congress to secure rights to “Authors” and “Inventors,” which is why modern copyright and patent doctrine remain focused on whether a natural person supplied the legally relevant creative or inventive contribution. In Burrow-Giles Lithographic Co. v. Sarony and Feist Publications, Inc. v. Rural Telephone Service Co., the Supreme Court treated copyright as requiring originality that comes from an author rather than from a merely mechanical process.
Trademark law is structured differently. The operative statutes are 15 U.S.C. § 1051 and 15 U.S.C. § 1052, together with the United States Patent and Trademark Office (USPTO) application regulations, and the central questions are whether a designation is distinctive, nonconfusing and used in commerce. In Two Pesos, Inc. v. Taco Cabana, Inc. and Jack Daniel’s Properties, Inc. v. VIP Products LLC, the Supreme Court again described trademark law in source-identifying terms, not in terms of authorship.
Copyright Asks the Most About AI Use
For copyrights, the statutory starting point is 17 U.S.C. § 102, which protects original works of authorship, and 17 U.S.C. § 201, which vests ownership initially in the author. The Copyright Office’s 2023 registration guidance makes the disclosure rule explicit for works containing AI-generated material.
That guidance says applicants have a duty to disclose AI-generated content included in a work submitted for registration, identify the human-authored material in the “Author Created” field and exclude AI-generated content that is more than de minimis in the “Material Excluded” section. The office also instructs applicants not to name an AI system or the vendor behind it as an author or co-author because the claim must be limited to human authorship.
The office’s 2025 Part 2 report on copyrightability sharpens the same point. It concludes that given current generally available technology, prompts alone do not provide sufficient human control over the expressive elements of the output while creative selection, arrangement and meaningful modification of AI-generated material may still support protection for the human-authored aspects of the final work.
That position now aligns with appellate authority. In Thaler v. Perlmutter, the D.C. Circuit held that the Copyright Act requires works to be authored in the first instance by a human being while also making clear that a human may still obtain protection for a work created with AI assistance if the legally relevant authorship is human.
For confidentiality purposes, copyright is the least private of the three systems. The registration record becomes public, and the office expressly explains in its privacy FAQ and public records portal that registration records are available for public inspection. But the compelled disclosure is still categorical rather than operational: Current guidance does not require applicants to reveal prompts, model settings, training data or the full workflow that led to the claimed work.
Patent Law Cares About Human Inventorship, not a General AI Confession
Patent law begins with 35 U.S.C. § 100(f), which defines an inventor as an “individual,” and 35 U.S.C. § 115, which requires an inventor’s oath or declaration. In Thaler v. Vidal, the Federal Circuit held that the Patent Act requires inventors to be natural persons.
The USPTO’s current Revised Inventorship Guidance for AI-Assisted Inventions does not create a separate inventorship test for AI. Instead, it says the same legal standard applies to all inventions, treats AI systems as tools used by human inventors, and states that the office generally presumes the inventors named on the application data sheet or oath are the actual inventors unless the record gives reason to do otherwise.
That is an important distinction for disclosure. The present patent statutes and guidance focus on who conceived the claimed invention, not on forcing applicants to identify every software tool involved in the inventive process. Put differently, current patent law does not impose a general requirement to list the model, the prompts or the vendor merely because AI was used somewhere in development.
What patent law does require is human responsibility for the filing. The USPTO’s 2024 AI tools practice guidance reminds practitioners and parties that existing rules already apply when AI is used before the office, including the duty of reasonable inquiry and the duty to ensure that filings have evidentiary and legal support. The same guidance also ties AI use to client confidentiality and warns that using outside AI systems for drafting, prior art searching or other tasks can expose client-sensitive information to third parties.
Patent practice also retains a meaningful zone of prepublication secrecy. Under 35 U.S.C. § 122 and 37 C.F.R. § 1.14, pending patent applications are generally preserved in confidence unless and until publication rules apply. So the patent system is not fully private, but it is still more confidential than copyright registration at the front end.
Trademark Law Is the Outlier
Trademark law does not ask who authored a mark, logo or brand name. It asks whether the applicant owns the mark, what goods or services are covered, what filing basis applies and whether the asserted designation functions as a source identifier. The current USPTO application rules in 37 C.F.R. § 2.32 list the required contents of a trademark application, and they do not require the applicant to explain how the mark was conceived or whether generative AI helped create it.
That means an AI-assisted brand name or visual brand indicium may still be registrable if it is otherwise distinctive and used properly in commerce. The disclosure question is largely absent because trademark validity does not turn on a human authorship inquiry in the way copyright does, or on a human conception inquiry in the way patent law does.
There is still a practical limit that matters. If the application is filed on a use basis, the applicant must supply a genuine specimen showing real marketplace use, and the USPTO’s specimen guidance stresses that the specimen is proof of what consumers actually see in commerce. So AI-generated mockups or idealized marketing images may create an evidentiary problem if they do not reflect actual use, even though AI provenance itself is not a disclosure item.
This divergence creates a striking possibility for the same visual asset. A logo generated with AI may be capable of functioning as a trademark without any AI-use disclosure to the USPTO’s trademark side, yet the identical image may have little or no copyright protection unless a human supplied protectable authorship or properly limited the claim in a copyright application.
What Can Remain Confidential
Across all three systems, current federal law still leaves substantial room to protect the process itself. Prompts, internal evaluation criteria, model settings, selection workflows and proprietary brand-generation or invention-development playbooks may qualify for protection under the Defend Trade Secrets Act if they derive value from secrecy and are actually kept secret.
But secrecy must be managed, not assumed. The USPTO’s AI tools guidance ties the issue directly to professional responsibility, and 37 C.F.R. § 11.106 and 37 C.F.R. § 11.18 underscore that client information and filing accuracy remain the practitioner’s responsibility, even when an AI system is used as an assistant.
That means the safest modern view is not that AI use is private by default. It is that the government usually wants the legal result of the human contribution identified while the operational details of the workflow can often remain confidential if the owner uses secure tools, contractual protections and internal controls strong enough to preserve trade secret status.
A Practical Rule for IP Owners
The simplest compliance rule is to separate the asset from the process. For copyright, document the human-authored contribution and be prepared to disclaim nonhuman output. For patents, identify the natural person or persons who actually conceived the claimed invention and do not let AI drafting obscure inventorship or filing accuracy.
For trademarks, focus less on how the name or logo was brainstormed and more on distinctiveness, clearance and authentic specimens of use. If a logo is valuable both as a mark and as a work of visual expression, evaluate trademark and copyright strategy separately because the same image can receive very different treatment in the two systems.
In the end, the most interesting point is not that AI has erased confidentiality. It is that AI has exposed how differently U.S. intellectual property law thinks about creativity itself. Copyright asks who authored expression, patent law asks who conceived the invention, and trademark law often asks only whether consumers see a source identifier.
Anton Hopen is a registered patent attorney and a shareholder at Trenam Law in Tampa, Florida. He is board certified in intellectual property by The Florida Bar. He can be reached at [email protected].
