NEWS
Copyright Office Report on AI Training and Fair Use
Can AI Use Copyrighted Material Under Fair Use?
Can AI systems rely on the fair use doctrine when training on copyrighted works? That question sits at the center of ongoing copyright debates.
Fair use is a doctrine in U.S. copyright law. It allows limited use of copyrighted material without permission in certain situations. Courts apply a four-factor test to determine fair use. Those factors include the purpose of the use, the nature of the work, the amount used, and market impact.
However, these factors are only guidelines. Importantly, no single factor guarantees fair use on its own.
Common Examples of Fair Use
Fair use often applies to criticism, commentary, and news reporting. It also commonly applies to teaching, scholarship, and research. For example, a reviewer may quote a copyrighted book without permission when critiquing that book. However, fair use is context-specific. Uses outside these categories may still qualify, while listed purposes may still fail.
AI Training Raises New Fair Use Questions
Generative AI systems train on massive datasets. Those datasets often include copyrighted material. As a result, a key question arises. Can AI training qualify as fair use and avoid infringement? This issue was the focus of a recent study. The study was issued by the U.S. Copyright Office (USCO).
Do AI models Infringe Training Data?
The USCO report concludes that AI training implicates several exclusive copyright rights. These include reproduction and derivative work rights.
The report also examines a critical issue. Are a model’s weights themselves infringing copies?
Developers argue that models contain only numerical values. They claim these values are not copies of copyrighted works. Others disagree, pointing to AI outputs that closely resemble training data.
According to the USCO, similarity matters. When outputs are substantially similar, infringement concerns grow stronger. In those cases, the report finds a compelling argument. Copying model weights may implicate reproduction and derivative rights.
How the Fair Use Defense Applies to AI
The key issue when analyzing fair use has been whether the use is “transformative”. There is of course a spectrum. Where the output is based on a diverse dataset, it is more likely to be transformative. But where it is trained to generate outputs substantially similar to copyrighted works, then it is “at best, modestly transformative”.
Why the Human Learning Analogy Falls Short
The USCO rejected arguments that AI training is inherently transformative, because it is analogous to human learning. The report asserts that the analogy rests on a faulty premise. Fair use is not a defense for all acts if those acts are used for learning. An example was given that a student couldn’t rely on fair use to copy all of the books at a library. The analogy also breaks down technically. Humans retain imperfect, filtered impressions of works. AI systems do not. The structure of exclusive copyright rights is premised on certain human limitations.
What Comes Next for AI and Copyright
Currently, there are more than 40 cases currently pending related to the issue of AI using copyrighted materials. There won’t be a single answer whether unauthorized use is fair use. There is an ongoing discussion of some form of licensing scheme for AI training data. For now, the recommendation is to continue development without government intervention.
More copyright, AI, and Fair use articles: AI and Copyright: Court Rules Against Fair Use in Training AI Models
Revised Policy on Same-Day Dual Filings in China: Invention and Utility Model Applications
China's Two Patent Model Types
There are two types of patent protection in China: invention and utility model patents. Invention patents are similar to the US utility patents. Whereas, utility model patents are granted for technical solutions relating to shapes or structures of a product.
Same Day Dual Filing
Under Chinese patent law, when the same applicant files both an invention patent application and a utility model application for the same invention on the same day (i.e., same filing date), it is considered a same-day dual filing. Whether these applications relate to the same invention is determined based on a declaration made in the request form.
To obtain a patent for the invention, the applicant must waive the corresponding utility model right. The older practice of amending the invention application to distinguish it from the utility model is no longer permitted.
Now You Must Choose One
Previously, if the utility model right was still valid and a declaration was submitted at the time of filing, applicants had two options:
- Modify the invention application to differentiate it from the utility model, or
- Waive the utility model right.
Now, only the second option—waiving the utility model right—is allowed. This amendment is intended to streamline examination procedures and reduce the burden on both applicants and the patent office.
Implication for Applicants
At the time of filing in China, applicants should carefully assess:
- Whether filing for both invention and utility model is necessary, especially if the scopes of protection are the same.
- If the scopes are identical, whether to pursue protection via a utility model or an invention patent.
Other articles on Chinese Patent Policy: Update on the Chinese Foreign Filing License Requirement
What Did Those Claims Consist Of?
This is a very weird case interpreting “consisting of”. The Federal Circuit recently affirmed a ruling of non-infringement in Azurity Pharmaceuticals v. Alkem Laboratories. Azurity had sued Alkem for allegedly infringing patent claims related to a drinkable antibiotic.
The patent at issue—Azurity’s U.S. Patent No. 10,959,948—was a continuation of an earlier application that had been rejected due to prior art referencing the ingredient propylene glycol. To overcome that rejection, Azurity amended its claims to disclaim propylene glycol, arguing that its absence distinguished the invention from the prior art. Instead, Azurity used the more generic term “flavoring agent.”
How the Prosecution History Set the Trap
During prosecution, Azurity clearly and unmistakably argued that its claimed formulation excluded propylene glycol in order to overcome prior art. That argument worked—at least to secure allowance.
But prosecution disclaimer cuts both ways. Statements made to obtain a patent can later limit its scope. By distinguishing prior art based on the absence of propylene glycol, Azurity boxed itself in.
Why “Consisting Of” Left No Wiggle Room
In patent law, wording matters.
- “Comprising” is open-ended. It allows additional, unrecited elements.
- “Consisting of” is closed. It excludes anything not specifically listed.
Here, the claims used “consisting of,” which strictly limits the formulation to only the listed ingredients. When paired with the prosecution history disclaiming propylene glycol, the claim language reinforced a narrow construction.
The Stipulation That Didn’t Save the Claims
At trial, both parties stipulated that “flavoring agents” in the asserted claims could include substances with or without propylene glycol. Azurity argued that this meant a product containing a flavoring agent with propylene glycol would still infringe.
The district court disagreed. A litigation stipulation cannot erase a clear and unmistakable prosecution disclaimer. Even if “flavoring agent” could theoretically encompass propylene glycol, the intrinsic record controlled.
Federal Circuit Affirms Non-Infringement
The Federal Circuit agreed: Alkem’s product, which contained propylene glycol, fell outside the scope of claims that—through both language and prosecution history—excluded it.
The takeaway is simple but powerful. When you amend claims to overcome prior art—especially when using closed language like “consisting of”—you may be defining not just what your invention is, but what it can never cover.
For some other articles about patent claims: Patent Like You Mean It, But Also Show the Structure
Federal Circuit Clarifies Limits on Prosecution Disclaimer Across Patent Families
Mission Impossible: Trademarking the US SPACE FORCE
Overview of the US SPACE FORCE Trademark Dispute
In a recent ruling, the U.S. Court of Appeals for the Federal Circuit upheld a decision by the Trademark Trial and Appeal Board (TTAB) to deny registration of the trademark “US SPACE FORCE.” The case offers a useful look at how trademark law intersects with national identity and public perception.
How the Space Force Name Entered the Public Domain
The term “Space Force” entered the public spotlight in March 2018, when then-President Donald Trump announced plans for a new military branch focused on space operations. By December 2019, the United States Space Force was officially established as the sixth branch of the U.S. Armed Forces.
The Timing of the Trademark Application
Just days after Trump’s 2018 announcement, intellectual property attorney Thomas Foster filed an application to register the trademark “US SPACE FORCE” for a range of goods and services on behalf of his law firm.
The timing was strategic. However, it ultimately proved insufficient to secure trademark rights.
Why the USPTO Rejected the US SPACE FORCE Trademark
The United States Patent and Trademark Office (USPTO) initially refused the application under Section 2(a) of the Lanham Act. That provision bars registration of marks that falsely suggest a connection with a person, institution, belief, or national symbol.
Section 2(a) of the Lanham Act and False Association
The USPTO concluded that “US SPACE FORCE” would lead consumers to believe the mark was associated with the United States government. As a result, the Office found the mark unregistrable.
The Appeal Process Before the TTAB and Federal Circuit
Foster appealed the refusal to the TTAB, which affirmed the USPTO’s decision. He then sought reconsideration, which was denied. The dispute ultimately reached the U.S. Court of Appeals for the Federal Circuit.
When Trademark Rights Are Evaluated Under Federal Law
Foster’s primary argument focused on timing. He asserted that the trademark should be evaluated based on the facts existing at the time the application was filed, not on later developments.
Filing Date vs. Examination Date in Trademark Review
The Federal Circuit rejected this position. The court held that the proper time to assess whether a mark falsely suggests a connection is during examination, not at filing.
By the time the USPTO examined the application, the government had clearly signaled its intent to use the name for a new military branch.
The Four-Part Test for False Suggestion of Connection
Courts apply a four-part test to determine whether a mark falsely suggests a connection:
- The mark is the same as, or a close approximation of, a name or identity previously used by another person or institution.
- The mark would be recognized as pointing uniquely and unmistakably to that person or institution.
- The named person or institution is not connected with the applicant’s activities.
- The person or institution’s fame or reputation is such that a connection would be presumed.
Why US SPACE FORCE Failed the False Association Test
In this case, all four factors were satisfied. The phrase “US SPACE FORCE” clearly pointed to a well-known government entity. That entity had no connection to Foster’s law firm or its services.
Government Names and Unmistakable Public Association
The court emphasized that a mark does not need to exactly match an official government name to create a false suggestion of connection. Even though the Space Force was not formally established at filing, the name was unmistakably tied to the United States.
Why Pop Culture References Did Not Matter
Foster argued that prior fictional uses of “Space Force,” including television references, weakened the association. The court disagreed. The strong public linkage to a newly announced military institution outweighed any fictional or generic uses.
Key Trademark Law Lessons From the Space Force Case
This decision highlights how timing, context, and public perception shape trademark rights. When a mark closely aligns with national identity or government authority, registration becomes far more difficult. For businesses and attorneys alike, the case reinforces the need to consider not only what a mark says, but what it signals to the public.
Related article on Trademark principles: A Tale of Trademark Law, the Rogers Test, and Parody
Decluttering the USPTO’s Trademark Register
The USPTO recently announced it had now removed over 50,000 goods and services from the trademark register. Using ex parte expungement and reexamination proceedings, the majority of the cancellations were used against registrations linked to specimen farms.
Trademark registration requires that marks be used in real-life commerce. Specimens of the marks must be provided to the USPTO to demonstrate commercial usage prior to registration. This would serve to show a legitimate connection between the mark and the goods and services provided.
Specimen farms are websites designed to look like commercial sites. These sites list the products and the marks, which are used as specimens provided to the PTO. But the sites don’t actually sell the product.
There are filing firms, companies that assist in filing trademark applications, that created these e-commerce sites purely for the purpose of creating fake specimens.
Clearly the registry of these unused goods and services helps open up the path for legitimate companies and trademarks.
Articles on the USPTO’s ongoing efforts to clean up the registry: 52k more fraudulently filed trademark applications and registrations cleared
Patent Like You Mean It, But Also Show the Structure
What Is Means-Plus-Function Claiming?
Means-plus-function claiming is a way of writing a patent that describes what something does, not how it does it. For example, describing a machine that opens a jar, rather than describing all the parts—a motor, a gripper, and how they work together—you just say: “A means for opening a jar.” This is to say you want to protect whatever part does the opening of the jar, without naming the exact part.
Why Structure Still Matters in Functional Patent Claims
However, the actual parts, the structure, must be described elsewhere in the patent. This is a way to write less in the claims, but the downside is that if it doesn’t describe how it works well enough, the claim might be invalid.
Section 112(f): The Legal Framework for Means-Plus-Function Claims
Section 112(f) addresses means-plus-function claiming. The courts apply a two-step analysis for such claims.
The Two-Step Court Analysis Under § 112(f)
First, the court determines whether the claim limitation is drafted in a means-plus-function format, which would invoke Section 112(f). Second, if it is, the court identifies what structure, if any, disclosed in the specification corresponds to the claimed function.
Claim Indefiniteness and Section 112(b)
In cases lacking disclosed structure, the claim is invalid under Section 112(b), which deals with claim definiteness. In simple terms, a patent must clearly tell others what is being claimed so they know the boundaries of the invention.
Why Vague Functional Claims Risk Invalidation
Clear claims can be understood, competitors can avoid infringement, and courts can enforce the patent consistently. Vague or ambiguous claims can be ruled invalid for indefiniteness, which is especially relevant in means-plus-function claims.
If the patent application doesn’t describe the structure that performs the function, it may be invalid.
Federal Circuit Spotlight: Fintiv v. PayPal
At the heart of the Fintiv v. Paypal case were terms like “payment handler” and “payment handler service,” found in Fintiv’s asserted patents. These terms were intended to cover components that interface with various payment APIs.
However, the court found they lacked inherent structural meaning and instead described what the component does rather than what it is.
Functional Claim Language Without Structural Support
At the heart of the case were terms like “payment handler” and “payment handler service,” found in Fintiv’s asserted patents. These terms were intended to cover components that interface with various payment APIs. However, the court found they lacked inherent structural meaning and instead described what the component does rather than what it is.
Why “Payment Handler” Failed as a Structural Term
This triggered the court’s application of the means-plus-function analysis under Section 112(f), even though the word “means” was not used.
Why the Court Rejected Fintiv’s Arguments
Fintiv argued that “payment handler” had a recognized meaning in the art, citing dictionaries, examples from the patent, and even the Internet Open Trading Protocol. The court was unconvinced. It found that the terms merely recited functions—such as wrapping APIs or exposing common interfaces—without explaining how those functions were implemented.
The specification did not help. Instead of disclosing a specific algorithm or structure, it restated the functional goals. According to the court, that was not enough under Section 112(f). A person of ordinary skill in the art would not be able to implement the invention based solely on the description provided.
Lessons for Software Patent Drafting
Functional terms, even when they sound technical, can lead to invalidation if the specification does not back them up with structure. For software patents, that structure usually means an algorithm. Avoiding the word “means” does not automatically avoid means-plus-function treatment. Courts look at substance, not form.
Key Takeaway: Substance Beats Form in Means-Plus-Function Analysis
If a claim recites a function, the patent must explain how that function is performed. Without that disclosure, the claim risks being found indefinite and invalid.
Related articles on patent claims: Federal Circuit Reverses ITC Decision, Strengthening Patent Eligibility for Composition Claims
What Did Those Claims Consist Of?
Also of interest: FishFAQ 004 Words are Limitations
Who Gets Credit When AI Invents? A Look at the History—and Future—of Inventorship
For centuries, governments have tried to encourage innovation by offering inventors something valuable in return: ownership. The earliest forms of state-recognized intellectual property—like patents—were meant to reward people for coming up with new solutions to problems. But it hasn’t always worked the way we’d like to imagine.
From Royal Favors to Inventor Rights
Between the 14th and 17th centuries, the English crown handed out patents. But these weren’t always about protecting brilliant ideas—sometimes they were just political favors, granting monopoly rights to the well-connected.
Things began to shift in the 18th century. Under Queen Anne, inventors had to submit detailed specifications explaining how their inventions worked. This marked a move toward recognizing patents not as tools of privilege, but as protections for genuine intellectual contributions.
In the U.S., this principle was written into the Constitution.
Queen Anne
The Patent and Copyright Clause gave Congress the power to protect inventors’ rights “to promote the progress of science and the useful arts.” James Madison himself argued that useful inventions belonged to those who created them. By the late 1800s, several countries—including England, France, and Germany—had formal patent laws recognizing inventors’ intellectual ownership.
The Core of Modern Inventorship: Conception
Today, especially in U.S. patent law, inventorship hinges on a key concept: conception. The inventor isn’t necessarily the person who builds the invention, but the one who conceives it—the one who had the original idea.
Some ground rules:
- Only someone who contributes to the mental formulation of the invention can be an inventor.
- There can be multiple inventors, but each must contribute to at least one claim in the patent.
- Simply carrying out instructions or building a prototype? That’s not inventorship.
Over time, the legal system evolved from handing out monopolies to truly honoring the individuals behind technological progress—those whose ideas shaped the future.
Then Came Generative AI
In recent years, AI has complicated things.
Autonomous AI systems can now generate ideas, designs, and even prototypes. But who’s the inventor when a machine contributes core ideas? Most global patent offices, including the U.S. Patent and Trademark Office (USPTO), say only a natural person—a human—can be listed as an inventor.
The courts have backed this up. In Thaler v. Vidal, the Federal Circuit ruled that AI can’t be an inventor under current U.S. law. However, that ruling left a critical question open: What about inventions developed by humans working with AI?
The USPTO's 2024 AI Guidance
In early 2024, the USPTO released new guidance aimed at clarifying this gray area. The agency said that a human can be listed as an inventor of an AI-assisted invention if they make a “significant contribution” to the final invention—even if they didn’t fully conceive it.
This marks a shift. Under this standard, a person could take AI-generated output, shape it meaningfully, and claim inventorship—even if their input alone wouldn’t qualify as an invention.
Legal Risks on the Horizon
This softer standard raises eyebrows in the legal community. Critics argue it contradicts patent law’s long-standing requirement of full human conception. And because USPTO guidance doesn’t carry the force of law, patents issued under this standard might not hold up in court.
There’s also risk for patent attorneys. Filing patents based on this weakened inventorship standard could raise ethical issues under the duty of disclosure rules, especially if the AI did most of the heavy lifting.
And here’s a surprising twist: the USPTO’s standard for inventorship may now be lower than the standard for patentability itself.
Looking Ahead
We’re at a crossroads. On one hand, AI is a powerful tool for accelerating innovation. On the other, our current legal framework still centers on human inventorship.
Without new legislation or clearer judicial rulings, this tension will likely grow. For now, inventors, companies, and patent attorneys should tread carefully. The future of inventorship in the AI era is still being written—and it’s far from settled.
Check out our FishFAQ video: The History of Patents
For other articles related to AI and Inventorship: Federal Circuit to AI Innovators: Generic Machine Learning Isn’t Enough for a Patent