As the prevalence of artificial intelligence has grown significantly over the last few years, the United States Patent and Trademark Office (USPTO) has taken measured steps toward its implementation within the intellectual property ecosystem. The USPTO has consistently recognized the tremendous potential for artificial intelligence “to support the agency’s mission,” yet it has exercised extreme restraint as it continues to evaluate its implementation among intellectual property practitioners, inventors, and the agency itself.
For example, in October 2020, the USPTO published its report titled Public Views on Artificial Intelligence and Intellectual Property Policy, marking one of the agency’s earliest inquiries into the intersection of artificial intelligence and the existing intellectual property framework, as well as the “reliability of IP rights for emerging technologies, such as AI.” Similarly, the USPTO previously established an Artificial Intelligence and Emerging Technologies Partnership “to gather public feedback through a series of sessions on topics related to AI and innovation[.]” Additionally, efforts by the USPTO and several other governmental agencies preceded and supported President Biden’s October 30, 2023 Executive Order titled “Safe, Secure, and Trustworthy Development and Use of Artificial Intelligence,” which encapsulated the federal government’s perspective regarding the competing advantages and risks associated with the use of artificial intelligence:
“Artificial intelligence (AI) holds extraordinary potential for both promise and peril. Responsible AI use has the potential to help solve urgent challenges while making our world more prosperous, productive, innovative, and secure. At the same time, irresponsible use could exacerbate societal harms such as fraud, discrimination, bias, and disinformation; displace and disempower workers; stifle competition; and pose risks to national security.”
The landscape surrounding the federal government’s and the USPTO’s implementation of artificial intelligence continues to evolve; however, the principles guiding this evolution remain consistent. Since President Biden’s Executive Order, the USPTO has issued ongoing guidance emphasizing its continued concern with the reliability, accuracy, and trustworthiness of AI-based programs. For instance, the USPTO has repeatedly issued guidance—at times, warnings—regarding practitioners’ use of AI tools in proceedings before the Patent Trial and Appeal Board (“PTAB”) and Trademark Trial and Appeal Board (“TTAB”). These changes align not only with the agency’s prior findings and actions to combat risks posed by artificial intelligence to the inventive process but also with methods of addressing growing concern regarding the use of AI within the legal community, such as the American Bar Association’s initiatives “to provide the legal community with insights for developing and using AI in a trustworthy and responsible manner.”
The USPTO’s recent implementation of AI-based patent examination tools, “SimSearch” “DesignVision,” reflects both the agency’s holistic assessment of AI’s benefits and risks and the agency’s intent to maintain safeguards while enhancing its examination process through use of AI. For example, as an enhancement to the existing Patent-End-to-End (PE2E) search function, SimSearch allows examiners to input selected portions of a pending application as a query to facilitate a prior art searches amongst both “domestic and foreign patent documents.” Similarly, DesignVision allows examiners to “facilitate[e] prior art searching using images as a query” to simultaneously search multiple foreign and domestic registries via a single search interface.
SimSearch and DesignVision represent some of the USPTO’s initial efforts to tailor artificial intelligence tools for internal use, while also advancing its broader goal of preserving the United States’ position as a global leader of innovation. However, as in many industries, significant concerns remain regarding the reliability of work product generated by or reliant on AI-based programs. The USPTO’s implementation of SimSearch and DesignVision, along with its internal guidance on their use of such programs during the examination process, reflects this persistent concern, as highlighted in recent USPTO announcements:
“SimSearch is intended to augment—not replace—other search tools and is available to examiners through our internal search tool, Patents End-to-End (PE2E) Search.”
“DesignVision will augment—not replace—design examiners’ other search tools. Examiners can continue using other PE2E search tools and non-patent literature when conducting their research.”
While this consistency between the DesignVision and SimSearch press releases may appear minor, it reflects the USPTO’s clear stance regarding artificial intelligence. Although the implementation of these tools contributes the agency’s goal to “harness AI to support the agency’s mission,” their use—i.e., to only “augment” rather than replace examiner’s efforts—demonstrates the USPTO’s core belief that, in its current form, artificial intelligence cannot substitute the integrity, reliability, and judgement of its human workforce. Through this approach, the USPTO has reaffirmed its continued commitment to exercising caution in adopting AI within the intellectual property ecosystem.
For any inquiries related to intellectual property, reach out to us as info@thip.law.
Denmark is set to introduce new deepfake legislation aiming to grant its citizens copyright protection over their own faces, voices, and other personal attributes, according to a recent New York Times article.
The goal: make unauthorized and potentially hard-to-spot AI-generated deepfakes illegal to create and/or distribute.
A deepfake is a photo, video, or audio recording which has been altered or manipulated, often to misrepresent an individual for deceptive purposes. Most commonly created using AI models trained on large datasets of real recordings and/or images, deepfakes have been used to replicate facial expressions, voice patterns, and other natural behaviors. Such creations pose significant risks, including misinformation, reputational harm and defamation, promoting non-consensual pornography, fraud, identity theft, and other threats that erode trust in the media. These concerns are especially relevant given the global rise of AI, which undeniably poses a serious threat to personal identity.
Denmark’s novel proposal may aid to fill gaps in existing legal frameworks which have difficulty addressing the misuse of personal identity in the era of generative AI. At the same time, the potential adoption of the proposed legislation raises questions regarding whether the U.S. lags in addressing these issues, as it currently lacks comprehensive federal legislation to regulate deepfakes.
At the heart of Denmark’s proposed deepfake legislation is a plan to hold online platforms accountable by requiring prompt removal of harmful deepfakes or subjecting websites to penalties under EU digital regulations—an initiative that has already gained broad political support and is expected to advance later this year.
The proposal specifically amends sections of the Danish Copyright Act to expand protections beyond not only performing artists and their artistic performances, but also to individuals seeking intellectual property protection against the unauthorized distribution of realistic, AI-generated imitations of their physical traits. This marks a significant shift, as it effectively treats a person’s likeness as copyrightable work, enabling individuals to assert copyright claims over their own identity. Under this framework, individuals would be empowered to issue takedown notices, initiate infringement proceedings, and seek damages. Importantly, the law includes fair use exceptions for parody, satire, and social criticism to protect freedom of expression.
The amendments to the Danish Copyright Act raise questions in terms of whether copyright can, or should be, used as a tool to protect personal identity. They also invite comparison to jurisdictions like the U.S., where free speech protections are stronger, and likeness rights are primarily governed by state-level “right of publicity” laws. And while the U.S. does offer name, image, and likeness protections in the form of these right of publicity laws, there are critical differences. Specifically, the right of publicity in the U.S. varies significantly by state, as there is no federal right enforcing the right of publicity and only applies when a person’s identity is used for commercial gain. By contrast, Denmark’s model would create automatic, nationwide copyright protections over one’s likeness—treating it as a form of intellectual property, similar to how creative works by artists and writers are protected in the U.S. under the Copyright Act. Moreover, the Danish proposal applies broadly to unauthorized uses, not just commercial exploitation, which is the primary focus of most state right of publicity laws.
While the Danish model appears promising, the U.S. would likely face significant challenges in adopting a similar, federal approach addressing deepfake issues. This is due in part to the absence of a federal right of publicity, meaning such a system would require new federal legislation. Additionally, any effort to implement these protections would need to navigate complex issues involving First Amendment rights and the limitations of U.S. copyright law, which currently protect creative works—not individual identities.
In an era marked by widespread cosmetic alterations like Botox®, fillers, and plastic surgery, questions arise relating to how one’s “likeness” is defined and whether Denmark has considered the possible ramifications of its legislation. In the U.S., many states’ right of publicity laws already protect individuals’ identities as recognizable in public. Specifically, some courts will evaluate whether the likeness is recognizably you as currently represented, meaning cosmetic alterations may negate certain rights of publicity, such as name, image, and likeness if your old likeness is no longer recognizable as you. Whether state right of publicity protections apply to an individual’s current appearance and past appearances is not clear, particularly because “likeness” is defined state-to-state, with some states offering broader protections that may cover AI-generated images resembling you, while others offer more narrow interpretations.
In the midst of evolving appearances, this begs the question: how would Denmark’s legislation apply to AI-generated images of pre-enhanced individuals whose current appearance differs today? Would the legislation permit individuals to hold onto rights to their old appearances, such that AI-generated images of one’s past-appearance is still protected? Considering the U.S. state-specific right of publicity framework may or may not protect an individual’s appearance during a specific point in time, some ambiguity exists around old vs. new appearances.
Under Denmark’s legislation, it seems unclear whether individuals will maintain a copyright-like right over their traits, regardless of when the likeness was captured. For now, the plain reading of the amendments suggest it won’t matter whether an AI-generated deepfake is based on your pre- or post-enhancement appearance or current look—any realistic imitation of your likeness will be protected.
While this is a thought-provoking law, clarification will ultimately be required. It will be interesting to see how this legislation progresses and whether the U.S. takes note how deepfake laws continue to unfold in other countries.
For your copyright questions, please email us at info@thip.law.
The Trump Administration is discussing a significant change to patent fees, according to a recent Wall Street Journal report. The Commerce Department are considering the addition of an annual tax to patent holders equal to 1% to 5% of the overall value of the patent. This tax would be in addition to the existing maintenance fees that are paid every 3.5, 7.5, and 11.5 years, after patent issuance.
The proposed tax could raise some concerns for patent holders.
First, it’s difficult to ascertain how valuable a patent is as each patent is unique and generally directed towards a different invention. Without a publicly available baseline to serve as a measure for the value of a patent, the value of the patent cannot be determined. For example, in the stock market, millions of identical shares in companies are traded daily and are therefore easy to value. Likewise, in the real estate industry, houses can be compared to similar houses with similar features in the same area that have recently sold in order to determine value. In contrast, individual patents are sold or licensed in private transactions, making valuation hard to determine.
Second, many patents are filed on inventions that never make it to market. In fact, these patents arguably have “negative value” because of the costs associated with them (e.g., maintenance fees). They bring in no revenue in the form of royalties or by protecting a particular product or line of business.
If the Trump Administration is successful with implementing this tax, patent owners will need to be strategic in how they manage their portfolio. Although it’s too early to know if this proposed tax will be implemented, and what consequences may arise, Thomas Horstemeyer will continue to follow this news. For questions and guidance on your patent, email us at info@thip.law.
On April 18, the Federal Circuit Court released its decision in Recentive Analytics, Inc., v. Fox Corp., Fox Broadcasting Company, LLC, and Fox Sports Productions, LLC. In Recentive, the patents claimed the use of machine learning to solve problems confronting broadcasters, such as the Defendants (collectively “Fox”), related to determining the scheduling of live events and optimizing network maps, which determine the programs or content displayed by a broadcaster’s channels within certain geographic markets at particular times. The Federal Circuit held that certain types of machine learning patents are per se patent-ineligible. However, the Court also noted that the decision does not apply to all machine-learning patents.
This decision arose from an appeal of a District Court decision on a motion to dismiss by Fox. In that decision, the District Court ruled that the four patents asserted by Recentive Analytics were not direct to patent-eligible subject matter. The Federal Circuit affirmed the decision of the District Court, stating that “today, we hold only that patents that do no more than claim the application of generic machine learning to new data environments, without disclosing improvements to the machine learning models to be applied, are patent ineligible under § 101.”
Many have questioned whether the holding in Recentive heralds the end of patenting machine learning related inventions. However, Recentive is consistent with previous decisions by the Federal Circuit on subject-matter eligibility. As an example, independent claim 1 of U.S. Patent No. 10,911,811 (one of the four patents asserted) recites:
A computer-implemented method for dynamically generating a network map, the method comprising:
receiving a schedule for a first plurality of live events scheduled to start at a first time and a second plurality of live events scheduled to start at a second time;
generating, based on the schedule, a network map mapping the first plurality of live events and the second plurality of live events to a plurality of television stations for a plurality of cities,
wherein each station from the plurality of stations corresponds to a respective city from the plurality of cities,
wherein the network map identifies for each station (i) a first live event from the first plurality of live events that will be displayed at the first time and (ii) a second live event from the second plurality of live events that will be displayed at the second time, and
wherein generating the network map comprises using a machine learning technique to optimize an overall television rating across the first plurality of live events and the second plurality of live events;
automatically updating the network map on demand and in real time based on a change to at least one of (i) the schedule and (ii) underlying criteria,
wherein updating the network map comprises updating the mapping of the first plurality of live events and the second plurality of live events to the plurality of television stations; and
using the network map to determine for each station (i) the first live event from the first plurality of live events that will be displayed at the first time and (ii) the second live event from the second plurality of live events that will be displayed at the second time.
In this claim, if one were to replace the bolded phrase “machine learning technique” with “human,” the entire method could be implemented by a person using pen-and-paper. The computer generally, and machine learning specifically, is only recited as the mechanism or vehicle by which the abstract idea is applied or implemented. Viewed another way, it could be said the claim anthropomorphizes the computer, which is often an indication that the claim may be too general to be patent eligible. It is therefore unsurprising that this claim was found to be subject-matter ineligible by the court, as well as claims with similar scope in the other asserted patents.
There are several conclusions for clients to consider moving forward.
First, inventions that use or rely on machine learning are still patentable. The Federal Circuit limited their holding only to patents that do no more than claim the application of generic machine learning to new data environments, without disclosing improvements to the machine learning models to be applied. So, the door is still open to machine learning patent applications that disclose, and claim, improvements to the machine learning models being used. Moreover, other tests for subject-matter eligibility, such as the machine-or-transformation test, are still good law and therefore offer alternative avenues for proving patent-eligibility.
Second, patents that simply claim the application of a business method or other abstract idea using a machine learning model are unlikely to be granted or to survive motions to dismiss in litigation. This is consistent with prior case law on subject matter eligibility. For example, in Alice Corp. v. CLS Bank Int’l, 573 U.S. 208 (2014), the Supreme Court stated that it is not enough to render a claim patent-eligible by stating an abstract idea and adding the words “apply it with a computer.” The Federal Circuit’s decision in Recentive is consistent with the holding in Alice – holding the claims that recite an abstract idea applied with a machine learning model to similarly be patent ineligible.
How significant these consequences from Recentive prove to be will remain to be seen and will probably be resolved through litigation over the next few years. We predict that machine learning patent applications will require a great degree of specificity in the claim language and the detailed description in order to be granted.
Thomas | Horstemeyer has robust experience in patent writing and patent prosecution. For questions and guidance, email us at info@thip.law.
One of our clients recently asked us to provide the Pros and Cons of filing multiple, more detailed patent applications to cover a particular technology as opposed to filing a single, more general patent application. An overview of our answer is provided below.
Patents can be used to dissuade and/or legally prevent competitors from copying a business’ products. If done correctly, this can help create an “economic moat” around the company’s business, which should result in the company being able to charge more for its services, resulting in higher profit margins.
Including more detail in patents will almost always result in stronger patent protection. The additional detail allows us to draft a stronger series of strategic fallback positions in case our broadest claims are rejected during negotiations with the patent office. This can significantly increase the chances of a particular patent application being allowed. These types of fallback positions are also often crucial at trial or in related settlement discussions since, if a patent’s broadest high-level claims are invalidated, more technically focused claims will often still cover the product or service at issue.
Having multiple patents that cover different detailed aspects of a particular technology also typically provides stronger patent protection than a single patent covering the relevant technology at a higher level. This is because it generally costs more to review and challenge more focused patents. Also, because the patent office fees and legal fees associated with challenging the validity of patents are incurred on a per-patent basis, it becomes significantly more expensive to attack a patent portfolio if multiple patents are involved. As an example, the patent office fees for attacking a patent via an Inter Partes Review (IPR) are currently over $50k, and the average total cost of an IPR is currently about $350k – $500k per patent, with no guarantee of a positive outcome.
Also, having multiple applications pending on related technologies can provide options for effectively switching patent examiners during patentability negotiations if those negotiations become difficult, which can help avoid the need to potentially appeal an incorrect decision. This can significantly reduce the overall time and cost associated with obtaining a patent.
A good patent portfolio can be used as a marketing tool for potential investors or purchasers of a company. If constructed properly, the patent portfolio should methodically highlight the company’s current proprietary technologies and provide a roadmap for the unique aspects of the company’s upcoming products and features.
Providing more technical detail and separating key inventive concepts into separate patent applications (generally with separate descriptive titles) helps emphasize the key features that the company has exclusive rights in. We have seen situations where a company’s IP has taken a back seat in the valuation process because it wasn’t presented effectively in negotiations. Having detailed and effectively categorized and labeled patents can help avoid this.
A strong patent portfolio can help defend a company from potential patent lawsuits in two ways. First, any disclosure of a particular technology made in a particular patent application should block competitors from patenting that technology. This reduces the overall risk of patent lawsuits. Second, having strong patents in place often prevents competitors from initiating patent infringement lawsuits since they may be facing counterclaims for patent infringement in response to initiating the lawsuit.
Including more technical detail in patent applications will help to more effectively block competitors from obtaining similar patents of their own. It will also generally produce stronger patents, which should provide an additional deterrent effect to potential patent lawsuits.
Patents can often be used to generate additional revenue through licensing. Having stronger, more detailed patents should increase the chances of licensing opportunities.
The most significant downside to providing additional detail in patent applications is cost, both in terms of legal fees and the cost of using internal resources to help prepare and review the various patent documents. With costs in mind, we would always recommend a steady, methodical, and balanced approach to developing a strong patent portfolio. We also do our best to reduce inventor time in reviewing patent documents.
Another downside relates to public disclosure. Because patents are public documents, in most situations, anything that we include in them will be made public. There is an often-overlooked option to have U.S. patent applications not publish unless the corresponding patent is granted, which can address this concern to some extent, but it’s always important to make sure that we’re comfortable publicly disclosing any information that we include in patent documents.
In summary, if funding is available and disclosing detailed information regarding the related technology isn’t an issue, we typically recommend filing multiple, more detailed patent applications to cover a particular technology. This approach typically provides a company with stronger offensive and defensive patent protection and – especially in the case of startup companies – can often have a meaningful impact on a company’s valuation.
Last month, the Federal Circuit released its en banc decision in LKQ Corporation v. GM Global Technology Operations, LLC. The following day, the USPTO released its updated guidelines for how examiners should evaluate design patents in view of the shift in the law. Although much ink has been spilled on the topic by patent bloggers, it will take some time to determine the ultimate ramifications of the decision.
The “executive summary” of the decision is that the Federal Circuit changed the law regarding obviousness with respect to design patents, throwing out the Rosen-Durling test that had been used for ~50 years and adopting the Graham factors used for utility patents in its place. We predict that this will make design patents both harder to obtain here in the US moving forward and also easier to invalidate design patents that have already been granted in a post-grant proceeding.
Procedurally, this decision arose out of an inter partes review (IPR) proceeding before the USPTO’s Patent Trials and Appeals Board (PTAB) involving a design patent. No ruling was made by the Federal Circuit on the obviousness of the design patent itself. Instead, the case was remanded to the PTAB for further proceedings in view of the decision. However, as a fundamental reinterpretation of the law, the case has immediate effect and general applicability – applying to all pending design patent applications as well as those design patents already granted.
For background, design patents in the US have been evaluated under a separate test than utility patents when it comes to obviousness – the Rosen-Durling test. That test required for obviousness rejections of a design patent that (1) the primary reference must be “basically the same” as the challenged design claim and (2) any secondary references be “so related” to the primary reference that the features in one would suggest application of those features to the other. This means that the bar for an obviousness rejection of a design patent is quite high under the previous Rosen-Durling test – almost as high as the bar for an anticipation (i.e., “novelty”) rejection since there is only a hair’s breadth between “the same” and “basically the same.” This is one of the reasons why design patents have historically been relatively easy to acquire compared to utility patents.
The holding of this case is that design patents must be evaluated using the same factors as utility patents (the Graham factors) when it comes to obviousness, making the standard for design patents consistent with KSR and other recent decisions. The Graham factors include identifying the scope and content of prior art (where analogous art had not previously been a consideration for design patents), determining the differences between the prior art and the design claim (casting aside the previous similarity requirement), the level of ordinary skill in the art (which had not previously been a consideration for design patents), and whether one would have been motivated to combine the prior art to create the same overall visual appearance (which had not previously been a factor). Although unstated, this holding may also overrule some recent Federal Circuit case law regarding what qualifies as prior art for a design patent (e.g., In re Surgisil), which were decided while the Rosen-Durling test was still the law for evaluating the obviousness of a design patent.
There are several potential consequences for clients to consider moving forward. How significant these consequences prove to be will remain to be seen, and probably resolved through litigation over the next few years.
First, design patents will likely be harder to obtain in view of the prior art. This is because the universe of available prior art to sustain a rejection is greatly expanded. Moreover, examiners are allowed to consider reasons for combining references beyond whether secondary references are “so related” to the primary reference that the features in one would suggest application of those features to the other.
Second, design patents will be easier to invalidate post-grant for the same reasons they will likely be harder to obtain. This is a sword that cuts both ways. On the one hand, clients can invalidate design patents asserted against them more easily, which means clients are at a lower risk from design patent infringement lawsuits generally. On the other hand, it would make it harder for clients to assert any of their design patents against potential infringers if they desired to do so. Whether this is a net positive or a net negative probably depends on the client and its litigation profile.