Listen to this post

On Sept. 29, the USPTO issued another memorandum as well as a new flowchart addressing applicant “subject matter eligibility declarations” (SMEDs) submitted to establish eligibility under Section 101. This is now the third time the office has issued guidance on SMEs in less than a year, which itself seems to highlight their importance to office leadership.

At first glance, the new eligibility flowchart may feel a bit unusual because it explicitly cites MPEP provisions that are about something else: evaluating declarations seeking to establish nonobviousness, e.g., for proving unexpected results or commercial success. Applicants hoping to use SMEDs effectively must therefore learn to navigate this borrowed doctrinal machinery.

So, what do those provisions actually mean in the eligibility context? And what do these provisions actually require of examiners in order to properly “weigh evidence” for eligibility purposes?

Translating the Standard

As a rough translation, the new SMED flowchart appears to ask three questions:

  1. Did the applicant prove an eligibility-relevant fact?
  2. Is that fact tied to what the claim actually requires?
  3. Is the expert explaining technology already disclosed or adding missing technology afterwards?

The first question comes from MPEP § 716.01(c): Factual propositions require factual support, and attorney argument is not evidence. Therefore, in the eligibility context, an expert’s statement that “claim 1 is patent eligible” gets little credit because it merely announces a legal conclusion. But testimony explaining how the technology operates, what improvements it achieves, what technical problem existed at filing, what a skilled artisan would understand from the specification, or why a claimed operation cannot practically be performed in the human mind can all supply actual evidence about the factual premises needed to decide the resulting legal conclusion. The September memorandum gives essentially those examples, and it states that a SMED can bestow factual gravity on propositions that might otherwise exist only as advocacy.

The second question is usually referred to as “nexus” in the context of obviousness. The new flowchart thus asks whether there is a “factually and legally sufficient connection” between the evidence and the claim, borrowing directly from MPEP § 716.01(b). Thus, evidence that a commercial system is faster or more efficient may have little value if the feature creating that improvement is absent from the claim. In other words, the technological improvement cannot be floating free of what the claim requires. 

The third question is stated in a way that may seem somewhat cryptic. The flowchart asks whether the declaration is “supplying information that was required to be present in the original disclosure,” with a “yes” pointing toward lower or no probative value. Here, “required” is best understood to mean information that had to be part of the application as filed for the applicant later to rely on that technical feature, relationship, or mechanism to establish patentability. In other words, a declaration may illuminate a disclosed invention and describe its advantages, but it cannot supply a missing invention itself. Thus, an expert declaration can explain why a skilled artisan would have recognized a technical consequence of an already-disclosed architecture, but it cannot introduce an undisclosed architecture years later as a basis for eligibility.

None of this means that every SMED necessarily reaches the weighing stage. The flowchart first gives examiners familiar ways to test whether the evidence deserves substantial weight, particularly whether the asserted technological fact has a sufficient connection to what the claim actually requires. That same issue has long been a battleground for declarations offered to establish secondary considerations of nonobviousness. But once the USPTO determines that the declaration is probative, tied to the claims, and not supplying missing disclosure, the flowchart moves to a different question: What evidence is actually being weighed against it?

What Does “Weighing” Actually Mean?

This is where the borrowed framework becomes more interesting. The flowchart itself directs examiners to MPEP § 716.01(d) when weighing the probative value of the evidence. That provision, citing the Federal Circuit in In re Eli Lilly & Co., states that facts established by rebuttal evidence must be evaluated against other evidence supporting the prima facie case and “not against the conclusion itself.” In other words, this framework emphasizes that proof is supposed to battle proof. As the Eli Lilly case put it, proof is not supposed to battle a legal conclusion already “set in concrete.”

That distinction would seem to have real consequences when applied to eligibility. Suppose an expert provides sworn, factually supported testimony that a claimed database architecture eliminates a particular reconstruction operation and therefore provides a technological improvement to how a system processes historical data. The examiner might have contrary technical literature, an applicant admission, statements in the specification, or some identified defect in the expert’s reasoning. These are all things that can be placed on the other side of the evidentiary scale.

But sometimes the examiner may have none of those things, in which case there may simply be no competing factual evidence to weigh. In the eligibility context, MPEP § 716 means that the office cannot simply choose to disagree with a competent factual showing made through a SMED. And to maintain an eligibility rejection, the office needs to explain why those facts are insufficient to establish eligibility rather than ignoring or disagreeing with them.  Otherwise, a legal conclusion can start masquerading as a factual counterweight. The distinction matters because acknowledging a declaration is not the same thing as weighing it, and disagreeing with an expert is not itself contrary evidence. If the office disputes an established technical proposition, the question turns to what factual basis exists for that dispute.

That practical problem may provide a clue about the USPTO’s increasing attention on SMEDs. The fact that the office has now provided three memos, further training, a detailed flowchart, and Technology Center points of contact for evaluating SMED evidence makes the practical meaning of “weighing” worth watching.

The more consequential question is: What happens after a competent factual showing survives the threshold inquiries?At that point, the office must analyze eligibility on that factual record rather than choosing to disagree with proven factual premises without identifying some contrary evidence or defect in the proof. If the declaration establishes that the claimed architecture improves computer functionality, the analysis should proceed on that established factual premise unless the record contains a factual basis to reject it. The ultimate eligibility determination remains a legal conclusion, but it is a legal conclusion reached from the factual record actually established—not from an immovable premise that the claim is ineligible. That is what it means to weigh the evidence. And that is why SMEDs have the potential to matter: they can force the eligibility analysis to rest on an articulated evidentiary record rather than on an unexplained intuition. Of course, the examiner could also identify a factual basis to disagree with the expert or identify a reason the declaration should not be considered probative in the first place. But what the examiner cannot do is simply devise a contrary version of the facts in order to preserve the same conclusion. 

Listen to this post

AI developers who train large language models on copyrighted code just cleared a significant legal hurdle. On September 16, 2026, the 9th Circuit affirmed dismissal of the plaintiffs’ claim in Doe v. GitHub, Inc. under Section 1202(b) of the Digital Millennium Copyright Act, holding that Copilot does not “remove or alter” copyright management information from an existing work when it generates code, because it creates a new work rather than reproducing a stored one.

What Happened

A group of programmers who published open-source code on public GitHub repositories alleged that Copilot and Codex sometimes reproduce their code without the attribution, copyright notices, or license terms that accompany the original files. They framed this as a violation of the DMCA’s copyright management information (CMI) provisions rather than as ordinary copyright infringement, likely because the DMCA allows statutory damages of up to $25,000 per violation.

The plaintiffs raised two distinct theories. Under the first, an “input” theory, GitHub and OpenAI supposedly stripped CMI from licensed code before feeding it into Copilot’s training data. The panel refused to consider this theory at all, finding that the plaintiffs had forfeited it by failing to correct the district court’s repeated statements that the case was not about training.

That left the second theory, the “output” theory. Copilot allegedly reproduces near identical or identical copies of a programmer’s code in response to user prompts, without the CMI attached to the original. The 9th Circuit agreed the plaintiffs had standing to bring this claim, since they plausibly alleged a substantial risk that their code would eventually be reproduced this way. But standing only got them in the door.

Why the Claim Failed on the Merits

The court’s reasoning turns on what it means to “remove or alter” CMI. Judge Eric Miller’s opinion holds that both verbs describe an affirmative act performed on something that already exists. You cannot remove or alter CMI from a work unless there was a copy of that work, with its CMI attached, in the first place.

Copilot, according to the complaint’s own description, does not retrieve and display stored files the way a search engine does. It identifies statistical patterns in the training data and generates a new output based on the most probable completion of a prompt. Because the resulting code is a newly created work rather than a reproduction of an existing one, the court held it never contained CMI to remove or alter in the first place.

The opinion is careful to note that this is not a strict identicality requirement. Minor cosmetic edits will not save a defendant who otherwise reproduces a protected work wholesale and strips its CMI. What matters is whether the output can fairly be described as a copy of an existing work missing its CMI, or whether it is a new work that never had CMI attached to begin with. On the facts alleged here, the court placed Copilot’s output in the second category.

The court also flagged the stakes of ruling otherwise. Treating ordinary substantial similarity between an AI output and a protectable work as enough to state a Section 1202(b) claim would let the DMCA’s enhanced statutory damages swallow traditional copyright infringement claims, which cap statutory damages at $30,000 per work rather than per violation. The panel declined to let a DMCA claim do that work.

What This Means for Your Business

This decision does not resolve whether AI-generated code or other AI outputs infringe copyright. The panel expressly said it was not deciding that question, and the plaintiffs’ breach of contract claims against GitHub are still pending in the district court. What this decision does is close off one aggressive route to statutory damages against AI developers, at least within the 9th Circuit and at least on an output theory framed the way these plaintiffs framed it.

For companies building or licensing generative AI tools that train on copyrighted or licensed material, the practical lesson is to document how a model actually generates its output, not just what it was trained on. The plaintiffs here lost in large part because their own complaint described Copilot’s process in terms that supported the court’s characterization of its output as a new work, so companies defending similar claims will want a clear, accurate technical record showing the model predicts likely completions rather than retrieving stored files. That said, this ruling is no shield against copyright infringement exposure generally. The court repeatedly distinguished the DMCA claim from a traditional copyright claim and noted that substantial similarity between an output and a protected work could still support an infringement theory, so that analysis remains a live risk entirely separate from CMI removal.

There is also a procedural lesson buried in how this case unfolded. The plaintiffs’ input theory, arguably a stronger fit for the DMCA’s text, never reached the merits because counsel failed to correct the district court’s understanding of the complaint early in the case. Any company litigating a novel theory should track how a court characterizes its claims at each hearing and object on the record when that characterization is wrong, since an uncorrected mischaracterization can forfeit a theory entirely.

Looking ahead, the DMCA claims may not be the last word in this dispute. With those claims narrowed, the plaintiffs’ breach of contract theories, based on the open-source licenses attached to their code, remain pending before the district court, and that may be where the more consequential precedent on AI training and output eventually comes from. It is also worth watching whether other circuits adopt the same reasoning. The panel leaned on a 5th Circuit decision reaching a similar conclusion, but acknowledged that print-media examples do not map neatly onto generative AI, so other courts may draw the line between a new work and a copy with CMI removed differently as more AI cases reach the appellate stage.

Key Takeaways

  • The 9th Circuit affirmed dismissal of the plaintiffs’ DMCA output theory against GitHub, Microsoft, and OpenAI, holding that Copilot’s generated code is a new work rather than a copy from which CMI was removed or altered.
  • The plaintiffs forfeited a separate “input” theory, based on CMI allegedly stripped during training, by failing to preserve it after the district court repeatedly said the case was not about training.
  • The plaintiffs did have standing to bring the DMCA claim, since a substantial risk of future reproduction without CMI was enough at the pleading stage.
  • The ruling does not decide whether AI-generated code infringes copyright, and the plaintiffs’ breach of contract claims tied to open-source license terms remain pending.
  • Companies building or defending generative AI tools should document how their models generate output, preserve every legal theory at every stage of litigation, and watch the contract claims for the next significant development in this case.
Listen to this post

In VDPP v. Volkswagen, the Federal Circuit made short work of rejecting VDPP’s argument that settlement licenses are somehow different from ordinary licenses for marking purposes. Following up on Seeing More Clearly: Patent Marking Compliance in the Wake of VDPP v. Volkswagen, this post turns to a question that has been percolating in the background of the VDPP discussion:

Can a patent holder sidestep the marking statute by structuring its settlement agreements as covenants not to sue rather than licenses?

The short answer is probably not — but the question is closer than it appears, and patentees have every incentive to test it.

The Marking Statute: Is “For or Under” the Hinge?

The marking statute’s language is what makes this question interesting. Section 287(a) conditions the recovery of pre-notice damages on marking by the patentee and by persons making, offering for sale, or selling patented articles “for or under” the patentee:

Patentees, and persons making, offering for sale, or selling within the United States any patented article for or under them, or importing any patented article into the United States, may give notice to the public that the same is patented… In the event of failure so to mark, no damages shall be recovered by the patentee in any action for infringement, except on proof that the infringer was notified of the infringement and continued to infringe thereafter…

The phrase “for or under” is doing a lot of work, and it may be the hinge on which the covenant-not-to-sue question turns. To understand why, start with the Federal Circuit’s treatment of the license/covenant distinction in the exhaustion context. 

Has the Crack in the Glass Already Filled?

In Jacobs v. Nintendo, a panel at the Federal Circuit distinguished between the covenant not to sue and the license grant in a settlement agreement for implied-license purposes. The license provision, the panel found, “goes much further by granting [the manufacturer] an affirmative right to engage in the manufacture and sale,” while the covenant not to sue “would have been fully sufficient” just to free the manufacturer of its own liability. That distinction mattered in Jacobs because the analysis turned on the parties’ intent regarding downstream customers.

But TransCore v. Electronic Transaction Consultants Corp. put that distinction to rest, at least in the context of exhaustion. Citing to the Supreme Court’s decision in Quanta that had made clear that the parties’ intent regarding downstream rights is irrelevant to patent exhaustion, TransCore clarified that what matters in the context of patent exhaustion is whether the sale was authorized. It held that the distinction between a license and a covenant not to sue is one of form, not substance:

[t]he real question… is not whether an agreement is framed in terms of a ‘covenant not to sue’ or a ‘license.’ That difference is only one of form, not substance — both are properly viewed as ‘authorizations.’

And, since a covenant not to sue authorizes sales just as effectively as a license does, the two seem to be functionally indistinguishable for exhaustion purposes.

The reasoning is grounded in the fundamental nature of patent rights. A patent does not give the patentee an affirmative right to practice the invention. Rather, it only gives the patentee the right to exclude others from practicing the invention. It thus follows that a patentee, by license or otherwise, can only convey a freedom from suit. The Supreme Court said as much in De Forest Radio: “[a]s a license passes no interest in the monopoly, it has been described as a mere waiver of the right to sue by the patentee.”

The Federal Circuit echoed this reasoning in TransCore:

A patent license agreement is in essence nothing more than a promise by the licensor not to sue the licensee. Even if couched in terms of ‘[l]icensee is given the right to make, use, or sell X,’ the agreement cannot convey that absolute right because not even the patentee of X is given that right.

The Nuance Is Real, Even If the Outcome Is Likely

While this question has been settled in the exhaustion context, it has not been squarely decided in the marking context. The VDPP panel left a sliver of hope, expressly declining to “foreclose the possibility that a licensor can ever establish it made reasonable efforts to ensure licensee compliance with 35 U.S.C. § 287 in the absence of a marking obligation.” A patentee might therefore press additional angles. An entity might start structuring settlement agreements as pure covenants — e.g., “we promise not to sue you for making, using, or selling products under patents X, Y, Z” — while explicitly disclaiming any license or authorization, and pairing this with arguments distinguishing this approach in the marking context from the prior decisions:

Textual Distinction.  Section 287 applies to “persons making, offering for sale, or selling within the United States any patented article for or under” the patentee. As the Federal Circuit explained in Arctic Cat Inc. v. Bombardier Recreational Products Inc., a patentee’s licensees must comply with § 287 precisely because the statute extends to persons making or selling patented articles “for or under” the patentee. But the marking obligation attaches to licensees, i.e., entities acting “for or under” the patentee, not merely to anyone the patentee has promised not to sue. A party operating under a bare covenant not to sue could argue it is not making or selling products “for or under” the patentee in the way a licensee does. It is simply free from the threat of litigation while acting under its own authority.

A pure covenant not to sue, unlike a license, does not affirmatively authorize the recipient to do anything. It merely removes the threat of litigation. The recipient acts under its own authority, not “under” the patentee. Under this theory, products sold by a covenantee would not be “patented articles” sold “for or under” the patentee, and the marking obligation would never attach. The “for or under” language arguably implies an agency-like relationship, i.e., someone acting with the patentee’s authorization or on the patentee’s behalf, and a covenant not to sue does not create that relationship. The statutory text in consideration here differs from the exhaustion framework, which turned on whether a “sale” was “authorized.”

Policy Distinction. The marking statute exists to provide constructive notice to the public that an article is patented. If a covenantee is not truly operating “under” the patentee’s authorization but is simply free from suit, the public-notice rationale is arguably weaker—the patentee has no meaningful control over the covenantee’s products and no relationship that would support imposing a marking obligation on the covenantee’s activities. 

The more likely outcome is that courts will treat covenants not to sue the same as licenses for marking purposes, consistent with the Federal Circuit’s well-established position that the two are functionally indistinguishable. Drafting around that position with a disclaimer is unlikely to work: The settlement language under consideration in TransCore itself disclaimed any “express or implied license,” and the court held the language did not change the result. Nor does the policy argument likely fare any better after VDPP since the panel held that excusing licensee marking would frustrate all three purposes of § 287 by putting products the patentee believes to infringe into the market unmarked.

But “likely” is not “certain,” and, as noted above, the question has not been squarely decided in the marking context. Patent holders structuring settlements should be aware that labeling an agreement as a “covenant not to sue” rather than a “license” is unlikely to avoid a marking obligation; the failure to mark could bar pre-notice damages altogether. On the other side of the table, accused infringers should scrutinize the patent holder’s settlement portfolio for exactly this kind of creative structuring, because it may open a marking defense that the patent holder assumed it had designed around.

Listen to this post

Is a pop star’s voice a trademark? Taylor Swift’s legal team is betting yes. On April 24, 2026, she filed three applications with the United States Patent and Trademark Office (USPTO) seeking federal registration of two sound marks and one image mark: the spoken phrases “Hey, it’s Taylor Swift” (Serial No. 99784980) and “Hey, it’s Taylor” (Serial No. 99784979), both spoken in Swift’s own voice, and a photograph of her performing on the Eras Tour stage in a sequined bodysuit with a pink guitar (Serial No. 99784977). The applications are filed in International Class 041 (entertainment services), rest on use in commerce dating back to at least 2011, and are supported by specimens pulled from Amazon Music and Spotify.

Unlike Swift’s prior trademark work protecting album titles and lyrics, these filings are not about merchandise or lyrics. They appear to be aimed squarely at AI-generated deepfakes, synthetic audio and video that can now convincingly mimic a celebrity’s voice and image without copying a single frame or note of anything she actually recorded.

Can You Really Trademark a Voice?

Registering a sound isn’t a new trick. MGM’s roaring lion, NBC’s three-note chime, and Intel’s audio bong have all held federal registration for years, and under the Trademark Manual of Examining Procedure § 1202.15, a sound qualifies as a mark once it is distinctive enough that a listener connects it to a single source rather than treating it as background noise or pure information. The bar climbs higher for anything commonplace (e.g. an alarm beep or a ringtone), which needs proof of acquired distinctiveness under Section 2(f) before the USPTO will register it. Zippo cleared that bar for the click-flip-click of its lighter by showing years of sales figures and advertising spend tied specifically to that sound.

What’s novel here is the subject matter — not a jingle, but a celebrity’s own spoken voice, introducing herself by name. That raises a real examination question, is “Hey, it’s Taylor Swift” functioning as a source identifier for entertainment services, or is it just Taylor Swift talking? Even if the mark registers, what is genuinely untested is whether it does any practical work against an AI-generated clip that merely sounds like her without reproducing the registered phrase word for word. The deeper doctrinal questions run further still. Courts have not worked out how to analyze likelihood of confusion when the allegedly infringing content comes from an algorithm rather than a person copying a phrase, or how liability might reach a platform or model developer that never itself displays the mark.

Swift isn’t the test case, though. In December 2025, the USPTO approved eight trademark applications from actor Matthew McConaughey covering his voice, image, and his signature catchphrase, “alright, alright, alright,” registered down to the pitch pattern of each syllable. McConaughey’s team was candid that their goal was to give him standing to sue in federal court over AI-generated content that mimics his voice or likeness, whether or not anyone is selling a product with it. If McConaughey’s registrations hold up, they are a strong signal the USPTO is willing to register these marks. The harder question is what they will actually be worth in an infringement fight.

Why Not Just Sue for Right of Publicity?

Celebrities already have a tool for unauthorized use of their name, image, and voice — the right of publicity. Bette Midler used it successfully against Ford in 1988 after the automaker hired her former backup singer to imitate her distinctive voice in a commercial (Midler v. Ford Motor Co., 849 F.2d 460 (9th Cir. 1988)), and Tom Waits won a similar case against Frito-Lay a few years later (Waits v. Frito-Lay, Inc., 978 F.2d 1093 (9th Cir. 1992)).

The catch is that right of publicity lives entirely at the state level. Coverage, remedies, and even whether the claim exists at all shift depending on which state’s law applies, and the doctrine grew up around human impersonators, not a model that can generate a synthetic voice on demand. A federal trademark registration sidesteps that patchwork. It applies nationwide and comes with causes of action built for exactly the kind of consumer confusion a convincing AI-generated “endorsement” would create. Registering these marks does not retire a right-of-publicity claim; it adds a federal option that can run alongside it.

Congress may soon add a third layer. The NO FAKES Act, which would create a federal property right against unauthorized AI-generated digital replicas of a person’s voice and likeness, advanced unanimously out of the Senate Judiciary Committee on June 22, 2026. If it passes, it will give artists (and everyone else) a federal statutory claim to sit alongside trademark and state right-of-publicity theories.

Not Just a Celebrity Problem

It is tempting to read this as pop-star news and move on, but the underlying problem is a business problem, not just an entertainment one. Companies build recognizable voices and images into their brands too, and those signals are just as exposed to synthetic imitation as Swift’s are. Disney didn’t wait for a celebrity test case to make this argument. In December 2025 it went straight after Google, demanding that the company stop its Gemini AI tool from generating images of copyrighted Disney characters, and Google reportedly took the material down within a day of receiving the letter. And Swift and McConaughey are not operating in isolation. Lionel Richie filed his own voice-and-likeness applications in June 2026, and the Backstreet Boys filed a sound mark for “hi, we’re the Backstreet Boys” that same month. What started as one artist’s novel filing is starting to look like a standard defensive playbook.

That playbook has real limits worth understanding before anyone assumes trademark registration solves the AI-impersonation problem outright. Training a model on existing recordings does not fall within trademark law’s reach. It’s not a commercial “use” of a mark in the sense the statute requires; it’s internal computation. A synthetic clip that merely sounds like someone, without using her actual registered phrase, sits outside the registration too. It is also only as strong as the paper trail behind it. Swift’s filings rest on a signed consent to register and a decade of documented use pulled from actual streaming specimens. And because no single doctrine covers the whole problem, the more durable answer tends to be a layered one: trademarks alongside state right-of-publicity law, contract terms addressing AI use, and, if the NO FAKES Act clears Congress, a federal statutory claim on top of all of it.

None of Swift’s three applications have registered yet, and there’s no guarantee they will. However, it is worth watching for the USPTO’s first office action on any of them, and for the day someone actually tries to enforce one of these registrations in an infringement suit rather than just filing it.

Listen to this post

Following up on What VDPP v. Volkswagen Could Mean for Patent Settlements and Marking and aspects of No Spectacle Worth Watching: How a Non-Practicing Entity’s Failure to Mind Its Licensees Cost It the Case, the Federal Circuit’s decision in VDPP v. Volkswagen makes clear that marking compliance is not a formality patentees can afford to view as optional, and settlement licenses do not get a pass.  Patent owners with licensing programs should revisit their agreements now, and accused infringers should treat a patentee’s licensing history as a first-order diligence item.

The Case at a Quick Glance

VDPP argued that licenses granted to settle litigation — especially where the licensee denies infringement — should not trigger the marking obligation under § 287(a). The Federal Circuit disagreed, holding that there is no substantive difference between a settlement-based license and any other patent license, particularly where, as here, each agreement was written as a standard license granting rights to make, use, and sell licensed products. 

The court further explained that a licensee’s denial of infringement does not matter because the marking inquiry focuses on the patentee’s own marking conduct.  In other words, a settlement license is a license for marking purposes, and a patentee that grants a license (including a settlement license) bears the burden of ensuring compliance with the marking statute for products made or sold under that license.

Failure to mark, or to ensure that licensees mark, limits the patentee’s ability to recover pre-notice damages.

Keeping Your Eye on the Mark: Steps for a Patent Owner

Below are some highlights from a more robust checklist. To obtain the complete checklist, please reach out here.

  • Audit existing license and settlement agreements. Review agreements for marking obligations (or their absence) particularly where licensees continue to sell covered products.
  • Add express marking obligations to future agreements, even where the licensee denies infringement. The court left open whether “reasonable efforts” short of a contractual marking clause can ever suffice, but offered no guidance on what those efforts might look like. Documentation of those efforts will therefore be important.
  • Verify licensing history against records and expect to disclose it if you file suit. Nondisclosure of relevant settlement agreements was among the factors that made this case “exceptional” for fee-shifting purposes.

What Accused Infringers Shouldn’t Overlook

  • Investigate the patentee’s licensing history early. A patentee’s prior settlement licenses — especially those lacking an express marking clause — can support a motion to dismiss a claim for pre-suit damages.
  • An NPE’s “nothing to mark” defense is not a complete answer. The obligation extends to any licensee selling a covered product. Look beyond the patentee’s own nonpracticing status to the conduct of its licensees.
  • Use the patentee’s own infringement position against it. Because a licensee’s denial of infringement does not defeat the marking obligation, a patentee’s continued assertion that licensed products practice the patent is itself evidence that marking should have occurred.
  • A successful marking defense can eliminate damages entirely, not just reduce them — particularly where the patent has expired and no post-notice infringement period remains.
Listen to this post

Trade secret lawyers commonly assert claims under both the federal Defend Trade Secrets Act (“DTSA”) and the applicable state’s version of the Uniform Trade Secrets Act (“UTSA”). But in doing so, they might overlook differences in the key elements and burdens of proof under those statutes. That can be costly, as shown by the Ninth Circuit’s recent decision in Comet Technologies USA, Inc. v. XP Power, LLC, 181 F.4th 988 (9th Cir. 2026): https://cdn.ca9.uscourts.gov/datastore/opinions/2026/07/14/23-15601.pdf.

In Comet Technologies, a jury instruction correctly tailored for the California Uniform Trade Secrets Act regarding the ascertainability of the alleged trade secret became reversible error after the plaintiff dropped its state law claims. With only federal trade secret claims remaining, the DTSA standard applied and shifted the burden from the defendant (who, under the California version of the UTSA, must prove the trade secrets were readily ascertainable) to the plaintiff, who under the DTSA must prove its trade secrets were not readily ascertainable. As a result, the appellate court overturned a $40 million verdict as well as a permanent injunction and more than $17 million in attorney fees.

Background

Three senior engineers left Comet Technologies, a computer chip component manufacturer, in February 2018 to work for competitor XP Power. They brought with them thousands of confidential files detailing Comet’s products, research and development strategy, and underlying technologies. They then delivered complete designs and a development plan to XP for new product lines in just nine days.

Comet sued XP under both the DTSA and California UTSA, asserting misappropriation of 20 alleged trade secrets. Comet narrowed its claims to five trade secrets and then, during trial, voluntarily dismissed its California UTSA claims to streamline the case. The jury found that XP misappropriated three trade secrets and awarded Comet $20 million in compensatory damages and $20 million in punitive damages. The district court later entered a permanent injunction and awarded Comet more than $17 million in attorney fees. XP appealed.

On appeal, the Ninth Circuit held that the district court improperly instructed the jury on the burden of proof regarding whether the alleged trade secrets were readily ascertainable, an essential element of Comet’s DTSA claims. The court reversed the $40 million verdict, vacated the permanent injunction and more than $17 million attorney fee award, and remanded for a new trial.

The “Readily Ascertainable” Burden Split

Information is “readily ascertainable by proper means” and thus is not a protectable trade secret if it could be lawfully “obtained, discovered, developed, reverse-engineered, or compiled without significant difficulty, effort, or expense.” Comet Techs., 181 F.4th at 994. This inquiry is purely hypothetical. It asks what a skilled competitor could have learned through proper means, not what the defendant actually did.

The appeal turned on which party must prove that Comet’s alleged trade secrets were “readily ascertainable through proper means.” Under the federal DTSA, lack of ready ascertainability is built into the statutory definition of a trade secret, and the plaintiff must prove that the information was not readily ascertainable. 18 U.S.C. § 1839(3)(B). Most states’ UTSA analogues, including Georgia’s (see Ga. Code Ann. § 10-1-761(4)), track the federal formulation and place the burden on the plaintiff.

Under the California UTSA, however, ready ascertainability is an affirmative defense, meaning the defendant must prove the information was readily ascertainable and acquired by lawful means. Cal. Civ. Jury Instr. No. 4420 (2025). While no other state’s UTSA includes California’s explicit burden-shifting instruction, other states, including Colorado, Illinois, and Oregon, have declined to adopt the DTSA’s “readily ascertainable by proper means” language in their statutory definitions of a trade secret (see C.R.S.A. § 7-74-102(4), 765 ILCS 1065/2(d), O.R.S. § 646.461(4)). The practical effect is similar to California’s UTSA on the front end: A plaintiff asserting trade secret misappropriation does not need to affirmatively prove non-ascertainability.

XP initially (and correctly) proposed the defendant-burden instruction when both Comet’s DTSA and California UTSA claims were pending. But once Comet dropped its California UTSA claims, the same instruction became erroneous. The Ninth Circuit rejected Comet’s defenses of the resulting error and found that (1) there was no invited error because XP timely and correctly objected, and (2) there was no showing that the error was harmless.

The damages methodology compounded the problem. Comet’s expert assumed that all confidential information was valuable to XP and calculated avoided research and development costs using internal cost codes. But because the expert provided a lump-sum figure for each trade secret, there was no way to isolate how much of the $20 million in compensatory damages turned on material that might have been readily ascertainable.

Practical Takeaways

  • Chart burdens claim-by-claim before trial. Pleading federal and state claims together is routine, but one should not assume they share the same burden structure.
  • Audit jury instructions and the verdict form each time claims are added, amended, or dismissed.
  • Build damages for appellate durability. Itemized, component-level proof is tedious, but a lump-sum number can present an “all or nothing” scenario and leave nothing for a reviewing court to salvage.
Listen to this post

The Federal Circuit’s decision in VDPP, LLC v. Volkswagen Group of America, Inc. offers a clear-eyed look at three recurring pain points in patent litigation: what it takes to plead, what is “exceptional” under 35 U.S.C. § 285, and what it takes to preserve an issue for appeal. The panel (Chief Judge Kimberly Moore, writing, joined by Judges Alan Lourie and Tiffany Cunningham) affirmed the dismissal of VDPP’s complaint without leave to amend, affirmed a six-figure attorney fees award against VDPP, and dismissed the appeal of sanctions against VDPP’s counsel for lack of jurisdiction. For an entity whose patent covers spectacles, VDPP proved remarkably shortsighted.

Peering into the Background

VDPP, LLC sued Volkswagen Group of America in the Southern District of Texas for infringement of U.S. Patent No. 9,426,452, which is directed to electrically controlled spectacles. Volkswagen moved to dismiss under Rules 12(b)(6) and 12(b)(3). VDPP asked for leave to amend and attached a proposed amended complaint, but the district court dismissed the case with prejudice and denied leave to amend as futile. But the court did not stop there. It later denied VDPP’s Rule 59(e) motion to change the final judgment, awarded Volkswagen $207,543.60 in attorney fees under 35 U.S.C. § 285, and sanctioned VDPP’s counsel, William Ramey, under 28 U.S.C. § 1927 and the court’s inherent authority, making VDPP and Ramey jointly and severally liable for the fee award. VDPP appealed all three rulings, and a fight over the sufficiency of the notice of appeal followed close behind.

Focusing on the Federal Circuit’s Analysis

The Federal Circuit’s opinion is a lens that magnifies every misstep by VDPP and its counsel.

Leave to Amend: An NPE with No Vision for Its Licensees’ Marking Obligations

Applying 5th Circuit law, the Federal Circuit reviewed the lower court’s denial of leave to amend the complaint for abuse of discretion, using the same legal-sufficiency standard that governs Rule 12(b)(6) motions. Because VDPP sought pre-suit damages, it carried the burden under 35 U.S.C. § 287(a) of pleading compliance with the patent-marking notice requirement, which applies to VDPP as well as any licensees. In this aspect, VDPP had previously entered into 11 settlement agreements licensing the ʼ452 patent, yet its proposed amended complaint asserted that it was a non-practicing entity with “no products to mark” and that “all conditions precedent to recovery are met.” The court treated those statements as bare legal conclusions, not factual allegations, and found nothing in the proposed pleading addressing whether any licensee had complied with § 287.

VDPP tried to argue that settlement-driven licenses are different in kind from ordinary licenses and therefore should not trigger the pleading obligation with respect to marking. The Federal Circuit was not persuaded, explaining that a license born of settlement is, in substance, no different from any other license, and that the accused infringer’s subjective understanding of the settlement is irrelevant to the § 287 analysis. Because every one of VDPP’s 11 agreements was framed as a standard license to make, use, and sell the licensed products, and VDPP continued to maintain that the licensed products infringed, the court saw no set of facts under which an amended complaint could plausibly allege that VDPP made reasonable efforts to ensure its licensees’ compliance. The complaint, in other words, never had the vision it needed, and no prescription was going to fix it.

Attorney Fees: A Pattern of Foggy Litigation Comes into Sharp Focus Under § 285

The Federal Circuit also affirmed the district court’s award of attorney fees to Volkswagen under 35 U.S.C. § 285, reviewing the exceptional-case determination for abuse of discretion. The district court had identified a laundry list of problems with how VDPP litigated the case including seeking future damages and injunctive relief on an already-expired patent, seeking past damages while unable to allege compliance with the marking statute, failing to disclose the very settlement agreements at the center of the marking dispute and making false statements about those agreements that prolonged the litigation. The lower court also flagged a pattern of “sloppy errors” and a broader history of repeat suits on the ʼ452 patent paired with settlement demands untethered to any real damages theory.

VDPP pushed back on several fronts, none of which gained traction. It blamed its nondisclosure of the settlement agreements on a misunderstanding traceable to its elderly president’s representations, but the Federal Circuit noted Volkswagen had told VDPP about those very licenses on multiple occasions well before that representation was made. VDPP also argued that conduct that would not independently support Rule 11 sanctions could not support a § 285 award, but the court reaffirmed the familiar Octane Fitness teaching that sanctionable conduct is not the benchmark for exceptionality; unreasonable conduct can be exceptional even if it is not independently sanctionable. And while a large volume of patent suits does not by itself prove an improper motive, the court held it was fair game for the district court to weigh VDPP’s pattern of litigation alongside the other evidence of unreasonable conduct in the record.

Sanctions: A Name Buried in the Fine Print Is Not a Clear Notice of Appeal

The most instructive piece of the decision from an appellate perspective may be the one the Federal Circuit never reached on the merits. Ramey wanted to challenge his personal sanctions, but the panel held it lacked jurisdiction to consider that challenge at all. Federal Rule of Appellate Procedure 3(c)(1)(A) requires a notice of appeal to specify the party or parties taking the appeal, and that requirement, though liberally construed, cannot be waived. Both of Ramey’s timely notices of appeal named only “Plaintiff VDPP, LLC” as the appellant; his own name appeared just once — misspelled — buried within a list of orders being appealed. Because the underlying order held VDPP and Ramey jointly and severally liable, the court found it was not clear from the notices that Ramey intended to appeal on his own behalf rather than simply on behalf of his client. His later “corrected” notices naming himself came more than 90 days after the appealed orders and were untimely under Rule 4.

Ramey leaned on a 5th Circuit case in which a notice of appeal was read to cover the attorney because the sanctions order there ran against the attorney alone. The Federal Circuit found that authority expressly limited by more recent 5th Circuit precedent to judgments naming only the attorney, not judgments like this one naming both the client and the attorney. VDPP separately tried to claim its own standing to challenge Ramey’s sanctions, but the court held VDPP suffered no injury from Ramey’s sanctions and, if anything, benefited from the joint-and-several arrangement. With no timely, adequately specific notice of appeal from Ramey and no standing for VDPP to stand in for him, the court dismissed that portion of the appeal for lack of jurisdiction, leaving the sanctions order undisturbed by default.

Key Takeaways

• A blind spot for a licensee’s marking requirements is damaging to damages. A patentee seeking pre-suit damages must plead compliance with 35 U.S.C. § 287, and that duty extends to licensees. Settlement-driven licenses get no exception, and a plaintiff’s own belief that licensed products infringe only sharpens the need to show the licensees marked.

• Enough scratches on the lens makes the whole view look exceptional. A district court may add up nondisclosure, misrepresentations, weak damages theories, and a pattern of low-value settlement demands, even where no single piece of that conduct would independently justify Rule 11 sanctions, to find a case exceptional under § 285.

• Don’t expect a court to read between the lines of a notice of appeal. Listing counsel’s name once, in connection with the order being appealed by the client, does not put anyone on notice that counsel is appealing on his own behalf. When a judgment holds a party and counsel jointly and severally liable, counsel’s own interests will not simply ride along on the client’s notice of appeal..

Listen to this post

On Tuesday of last week, the Federal Circuit gave Import Global a second chance in district court, reminding practitioners of the importance of careful claim language and a detailed written description. In Socket Solutions, LLC v. Import Global, LLC, the appellate court vacated and remanded a Florida district court’s grant of a preliminary injunction, concluding that the district court had misconstrued the claims and that, as a result, the likelihood of success required for injunctive relief had not been met.

Background

Socket Solutions, LLC, owns U.S. Patent No. 9,509,080 (the ʼ080 patent) — an indoor electrical wall outlet cover that permits functional use of an electrical wall outlet while fully concealing the plug contact openings of the outlet — and sells the patented product under the name Sleek Socket. At issue in this suit is claim 19 of the ʼ080 patent, which requires an apparatus for hiding a wall outlet while affording use of the outlet comprising (emphasis added):

  1. a cover comprising:
    • i. a frontplate; and
    • ii. a backplate comprising at least one set of electrical prongs including a hot prong, a neutral prong, and optionally a ground prong, positioned to correspond to a first receptacle of the wall outlet; and
  2. an electrical cord extending from the backplate, or the cover, said cord comprising at the cord’s proximal end: at least one hot pin, at least one neutral pin and optionally a ground wire positioned on or fastened or attached to the backplate of the cover in such manner as to minimize distance between the front plate and the backplate.

Alleging that Import Global infringed claim 19 of the ʼ080 patent, Socket Solutions sued and moved for a preliminary injunction. The motion was granted after determining Socket Solutions demonstrated a likelihood of success on the merits of its infringement claim. Import Global appealed, and the appellate court reviewed the grant of injunctive relief.

The Federal Circuit’s Analysis

In its review, the Federal Circuit focused on the district court’s construction of two terms as adopted from Socket Solutions’ recommendation: “backplate” and “pin.” The district court had construed “backplate” as the component of the cover, opposite the frontplate, that includes at least one set of electrical prongs. The Federal Circuit held that this construction erroneously imposed a spatial-reference requirement relative to the “frontplate” that the specification does not require. For the same reason, it declined to adopt Import Global’s proposed construction that the “backplate” is the portion of the apparatus closest to the wall outlet once the apparatus is plugged in. Relying on the written description, the Federal Circuit instead construed “backplate” as the component that, together with the frontplate, forms the cover, such that the cover’s maximum thickness is measured at its central portion between the frontplate and that component.

In agreement with Import Global, the Federal Circuit determined the district court incorrectly construed “pin” as a means-plus-function term. Under 35 U.S.C. § 112(f), a patentee may draft claims as a means or step for performing a specified function without the recital of structure, material, or acts in support thereof. When a claim term lacks the word “means,” there is a rebuttable presumption that § 112(f) does not apply. While that presumption can be overcome, the Federal Circuit held it remains intact where, as here, the written description further defines the claim term (“pin”) in structural terms.

Socket Solutions did not dispute that “pin” is a structure but argued that courts have construed structural terms in purely functional language without triggering § 112(f) when the structural element is best defined by its function. The Federal Circuit disagreed, reasoning that a purely functional definition would be overly broad. The appellate court reiterated that it departs from a claim term’s plain and ordinary meaning only through lexicography or disavowal, so, finding no prosecution history here suggesting otherwise, the appellate court gave “pin” its plain and ordinary meaning as understood by a skilled artisan.

Considering this determination, the Federal Circuit vacated the district court’s grant of a preliminary injunction and remanded for the court to reevaluate the likelihood of success of Socket Solutions’ infringement claim under the ʼ080 patent. The appellate court did not reach the merits of Import Global’s irreparable harm arguments.

Key Takeaways

  • Courts will not rewrite the claims. A court must look at the written description and patent prosecution to interpret claim language and will not add or omit limitations to reach a favored outcome, so claims should be written as they are intended to be interpreted.
  • Functional language doesn’t automatically trigger § 112(f). A broad claim term reciting a function is not converted into a means-plus-function limitation merely because the element could also be described functionally, so if a mean-plus-function is intended, the limitation should include “means.”
  • Courts will not read between the lines. Where spatial referencing is important to a product, it should be stated explicitly in the written description, as an appellate court is unlikely to infer product features on its own.
Listen to this post

In a decisive reversal, the Federal Circuit handed Boston Scientific Corporation a complete victory in Board of Regents of the University of Texas System v. Boston Scientific Corp., vacating a jury verdict that had found Boston Scientific’s drug-eluting coronary stent infringed the University of Texas’s patent for drug-releasing biodegradable fiber implants. The Federal Circuit concluded that Boston Scientific was entitled to judgment as a matter of law (JMOL) on both anticipation and noninfringement — wiping out the almost $50 million damages award in favor of the University of Texas (UT) and ending the case.

Background

U.S. Patent No. 6,596,296 (the ʼ296 patent) generally relates to compositions containing at least one biodegradable polymer fiber in which a therapeutic agent is dispersed. The ʼ296 patent is directed to the controlled delivery of therapeutic agents — including anti-inflammatory compounds and long-term cardiovascular drugs — via biodegradable fibers that can be used in conjunction with commercially available stents. Claim 1, the representative independent claim, requires a composition with at least one biodegradable polymer fiber made up of two immiscible phases, i.e., the polymer portion of the fiber and the discrete drug-containing regions dispersed throughout the fiber:

A composition comprising at least one biodegradable polymer fiber wherein said fiber is composed of a first phase and a second phase, the first and second phases being immiscible, and wherein the second phase comprises one or more therapeutic agents.

UT sued Boston Scientific in 2017 in the Western District of Texas alleging that Boston Scientific’s drug-eluting coronary stent systems infringed the ʼ296 patent. Boston Scientific’s accused stent consists of a metal frame containing linked, serpentine (zigzag-shaped) rings coated with a liquid drug-containing biodegradable polymer coating that dries like paint and adheres to the outer surface of the metal frame. As the coating biodegrades, the drug of interest is released over time.

The case was ultimately transferred to the District of Delaware where, after a two-phase trial, the jury found that Boston Scientific infringed asserted claims 1, 11, 17, and 26 of the ʼ296 patent, that the claims were not invalid as anticipated by U.S. Patent No. 5,364,627 (Song), and that Boston Scientific’s infringement was willful. As a result, the jury awarded damages to UT in the amount of $42 million (plus ~$7.4 million in pre-judgment interest). The district court subsequently set aside the willfulness finding as legally unsupported but otherwise upheld the verdict.

Boston Scientific appealed and UT cross-appealed the rejection of the willfulness finding.

The Appeal

On appeal, a central question was whether Song — which Boston Scientific’s expert characterized as disclosing a fiber-based drug delivery system — anticipates the asserted claims of the ʼ296 patent. Song discloses a delivery system for the gradual release of an active agent (such as a drug or chewing-gum flavorant) from a fiber, with the agent dispersed throughout the fiber and released upon exposure to a solvent. Yet, the district court had credited UT’s expert testimony that Song is effectively a “chewing gum” patent with a manner of biodegradation relevantly different from the ʼ296 patent. The Federal Circuit rejected this characterization across all four asserted claims. More specifically, with respect to the biodegradable polymer fiber limitation of claim 1, the Federal Circuit found Song’s express disclosures compelling, e.g., Song teaches a “fiber having a support matrix” made of biodegradable polymers. Indeed, the appellate panel found no reasonable basis for restricting Song’s teachings in the way UT proposed, i.e., limited to chewing gum applications.

On the two-phases limitation, the Federal Circuit found that Song’s express disclosure that the active agent “does not necessarily have to be in a contiguous phase” directly teaches the claimed “discrete drug-containing regions.” And, despite UT’s expert’s attempts to distinguish Song’s mechanism of drug release from the asserted claims — arguing that Song releases drugs through channels exposed by chewing and deformation as opposed the ʼ296 patent’s radially release of drugs — the court found this argument unpersuasive and beside the point. In particular, the appellate panel explained that the district court’s claim construction of “second phase” does not require any particular mechanism of release, so evidence about how the drug releases does not address the actual claim requirement. On the immiscibility limitation, the appellate panel noted that Song itself expressly states that the active agent and the fibrous wall material “must be immiscible with each other,” and UT did not separately argue otherwise.

With respect to the “varying rates” limitation in claim 26, the uncontradicted testimony of Boston Scientific’s expert established that all drug delivery systems release agents at varying rates because it is “very, very difficult” to manufacture a system where the release rate does not change with time. In this aspect, UT presented no expert testimony to the contrary and offered no persuasive counterargument at oral argument. The panel found that UT’s own trial submissions further supported the conclusion. In fact, during closing argument, UT’s counsel described the chewing mechanism as producing varying deformations and varying openings where solvent contacts the active agent, which the panel read as implicitly acknowledging varying rates. The panel also noted that Song’s figures depicted openings of various sizes at the ends and sides of a fiber, which affirmatively corroborated Boston Scientific’s expert.

While its anticipation conclusion was sufficient to resolve the appeal, the Federal Circuit panel also addressed noninfringement. The district court had construed “fiber” to mean “a thread-like structure of any length or shape.” UT’s infringement theory did not rest on the stent as a whole but instead focused on the polymer coating stripped from the metal frame (more specifically, the portion of the coating covering one serpentine (zigzag) loop around the stent). The appellate panel found that no reasonable jury could characterize that artificially separated section of stent coating as “thread-like” and further noted that the coating in Boston Scientific’s stent is not a structure that ever had or would have independent existence. Rather, the coating is the result of adhering a liquid to the frame and letting it dry such that it is not meant to be stripped from its frame or to retain its form on its own. The resulting detached coating section is tubular in overall shape with zigzag walls, not essentially one-dimensional like a thread but visibly two-dimensional like a ribbon. The panel drew a pointed analogy to further explain its finding:

[O]ne cannot reasonably say that a uniform sheet of paper contains a narrow strip simply because one could cut such a strip from the paper.

Notably, UT’s expert admitted that he had not at any point in the case opined that the polymer coating on Boston Scientific’s stents is thread-like. On that record, the appellate panel found that Boston Scientific was also entitled to JMOL of noninfringement.

As a side note, a threshold issue on appeal was the standard of review. Anticipation is a question of fact, and noninfringement based on claim construction is reviewed for substantial evidence when a jury has made the underlying factual findings. Under that deferential standard, the Federal Circuit would uphold the jury’s verdict so long as it was supported by substantial evidence. Here, however, Boston Scientific did not directly challenge the jury’s factual findings. Instead, it appealed the district court’s denial of its renewed motion for JMOL. Because the Federal Circuit applies the procedural law of the regional circuit in which the case was tried, the standard of review for the JMOL denial was governed by 3rd Circuit law, which reviews JMOL denials de novo. Under de novo review, the appellate court examines the record anew and asks whether the evidence, viewed in the light most favorable to the verdict winner, permits only one reasonable conclusion — the one favoring the movant. This procedural posture proved critical: De novo review gave the Federal Circuit substantially more latitude to reexamine and ultimately reject the evidentiary basis for both the anticipation and noninfringement findings than it would have had under substantial evidence review.

The result here is a clean sweep for Boston Scientific. In the end, the fiber in the polymer composition claimed in the ʼ296 patent could not hold the verdict together any better than it could hold its shape off the stent. 

Takeaways

This case serves as a stark reminder that even multimillion-dollar jury verdicts can unravel on appeal — and that the procedural vehicle chosen for the appeal can be outcome-determinative. The Federal Circuit’s decision offers important guidance on the strategic significance of the standard of review, anticipation analysis, the scope of expert testimony in invalidity disputes, and the limits of infringement theories built on artificial claim-element constructions:

  • The procedural vehicle for appeal can be outcome-determinative. Boston Scientific appealed the district court’s denial of its renewed JMOL motion rather than directly challenging the jury’s factual findings. Because the Federal Circuit applies regional circuit procedural law, this meant the appeal was reviewed under the 3rd Circuit’s de novo standard for JMOL denials — not the more deferential substantial evidence standard that would have applied to a direct challenge to the jury’s verdict. Under substantial evidence review, the jury’s findings on anticipation and noninfringement — which were supported by UT’s expert testimony — might well have survived. The de novo posture gave the Federal Circuit the latitude to independently assess the sufficiency of the evidence and conclude that no reasonable jury could have found in UT’s favor. Appellate practitioners should carefully consider whether the procedural framing of an appeal can unlock a more favorable standard of review.
  • Expert testimony that misses the claim requirement cannot rescue a verdict. UT’s expert testified extensively about how Song’s manner of drug release differed from the ʼ296 patent, but claim construction imposed no limitation on the mechanism of release. The Federal Circuit was unsparing: Testimony about an unclaimed aspect of the invention provides no legally sufficient evidentiary basis to support the verdict. Before trial, patent owners must carefully map every element of expert testimony to the actual construed claim language, not just to the broader inventive concept.
  • Prior art is what it says, not how it’s used. UT tried to pigeonhole Song as a “chewing gum” patent with a narrow field of application. The Federal Circuit rejected this framing because Song itself expressly discloses “drugs” as an active agent and clearly is not limited to gum or saliva as a solvent. When a prior art reference expressly discloses broader uses — here, drugs as well as flavorants — courts will not artificially narrow its teachings based on certain embodiments. Patent owners should not optimistically rely on a narrow characterization of the purpose of a prior art reference to avoid anticipation.
  • Uncontradicted expert testimony on anticipation creates real risk. On the “varying rates” limitation of claim 26, Boston Scientific presented expert testimony that went entirely unrebutted by UT. The one-sided record proved fatal to UT’s attempt to avoid anticipation. When faced with credible expert testimony on a specific invalidity ground, patent owners should ensure they put forward affirmative counterevidence at trial, not just legal argument on appeal.
  • Infringement theories must rest on the accused product as it actually exists. UT’s infringement case required the jury to imagine a portion of a stent coating stripped from its metal frame and treated as an independent thread-like structure. The Federal Circuit made clear that this kind of artificial reconstruction, which relied on a hypothetically detached section of a product that was never meant to exist independently, is not sufficient. Infringement theories must be grounded in the accused product’s actual structure and function, not in what a component might look like if removed and reimagined. Patent owners must make sure that the infringement theory matches reality.
Listen to this post

A new rule from the USPTO took effect July 20, effectively ending foreign pro se patent prosecution: Patent applicants and patent owners whose domicile is outside the United States or its territories must now be represented by a registered patent practitioner — and, as we’ll see, even some U.S. citizens are caught by it.

Here is a breakdown of who the rule affects, how it applies, and why the USPTO adopted it.

Who the Rule Affects

The rule applies to both patent applicants and patent owners. Per the USPTO, a patent applicant includes any inventor, joint inventor, legal representative, person to whom the inventor has assigned (or is obligated to assign) their rights, or anyone else who shows sufficient proprietary interest to be named as an applicant. A patent owner is the person or entity that holds title to an issued patent, as reflected in USPTO assignment records.

The rule turns on domicile — in other words, an individual’s permanent legal place of residence or a juristic entity’s principal place of business. For an inventor-applicant, domicile is normally determined by the residence information in the application data sheet (ADS) and/or the inventor’s oath or declaration. For a non-inventor applicant, it’s normally based on the mailing address in the ADS’s applicant information section. Where an application has multiple applicants or owners with different domiciles, the representation requirement applies to the group as a whole if any one of them is domiciled outside the United States.

This is not a citizenship test. The USPTO has confirmed it does not collect citizenship information, since doing so would create a significant risk and administrative burden. That means a U.S. citizen who is permanently domiciled abroad is still treated as a foreign pro se applicant and remains subject to the new requirement. That also means that a non-U.S. citizen permanently domiciled in the U.S. can file pro se.

It’s worth noting that juristic entities (corporations, partnerships, government agencies, trusts, and the like) were already required to be represented by a registered patent practitioner, regardless of domicile, when acting as an applicant. This rule extends that same requirement to juristic entities acting as patent owners.

What the Rule Requires

Those who qualify must be represented by a registered patent practitioner: a patent attorney or patent agent registered under 37 CFR 11.6, or an individual granted limited recognition under § 11.9(a), § 11.9(b), or § 11.16.

This requirement applies to all filings, including new applications, amendments, replies, and other papers received on or after the effective date, regardless of when the underlying application was originally filed. In practice, that means a foreign pro se applicant who filed before the rule took effect will still need a registered practitioner to respond to the office going forward in that same application. A reply filed without one may require corrective action, but new applications will still be accorded a filing date under the conditions set forth in 37 CFR 1.53.

The rule does not require practitioner representation for paying maintenance fees or the basic filing fee. It does, however, still require a practitioner’s signature to certify micro entity status for a reduced fee.

Why the USPTO Adopted It

According to the USPTO, requiring foreign applicants/inventors and patent owners to be represented by a registered patent practitioner:

  1. Treats foreign applicants/inventors and patent owners the same way U.S. applicants/inventors and patent owners are treated in other countries;
  2. Increases efficiency, since the USPTO spends significant resources assisting pro se applicants;
  3. Enables the USPTO to more effectively enforce compliance with statutory and regulatory requirements in patent matters; and
  4. Enhances the USPTO’s ability to respond to false certifications, misrepresentations, and fraud.

Consistent with other recent rule changes under Director John Squires, the final rule is also aimed at reducing patent application pendency: Pro se applications generally require additional processing by the Office of Patent Application Processing (OPAP) and additional examination time from patent examiners on procedural matters.

The Bottom Line

Any applicant or patent owner domiciled outside the U.S. will need a registered patent practitioner to file and respond to the USPTO. If the new rule affects you, Bradley has a number of experienced attorneys registered to practice before the USPTO who can step in on patent drafting, filing, and prosecution.