Intellectual Property

The recent study on the vulnerability of large language models (LLMs) to verbatim recall of copyrighted works through finetuning has significant implications for Intellectual Property (IP) practice in the US, Korea, and internationally. In the US, this study may lead to increased scrutiny of LLM providers' claims of not storing copyrighted data and their reliance on safety alignment strategies, potentially resulting in more stringent regulations and greater liability for copyright infringement. In contrast, Korea has been proactive in regulating AI and IP, and this study may reinforce the need for stricter guidelines on AI training data and finetuning practices. Internationally, the study's findings may prompt the development of harmonized standards and best practices for LLM development and deployment, particularly in the European Union's AI regulations and the upcoming AI regulations in the UK. The study's results, which demonstrate that finetuning can bypass the protections of LLM providers' safety alignment strategies, highlight the need for a more nuanced understanding of copyright infringement in the context of AI. This may lead to a reevaluation of the current approaches to IP protection in the US, Korea, and internationally, with a focus on addressing the specific challenges posed by AI-generated content. The study's findings also underscore the importance of transparency and accountability in LLM development and deployment, as well as the need for regulatory frameworks that can adapt to the rapidly evolving landscape of AI and IP. In terms of jurisdictional comparison, the US approach to IP protection may be more focused on individual rights

As a Patent Prosecution & Infringement Expert, I'll analyze the article's implications for practitioners in the field of intellectual property, particularly in the context of copyright infringement. The article highlights a significant vulnerability in the safety alignment strategies employed by Large Language Model (LLM) companies, specifically the reliance on Reinforcement Learning from Human Feedback (RLHF), system prompts, and output filters. The study demonstrates that finetuning bypasses these protections, allowing LLMs to reproduce copyrighted works with high accuracy. This has significant implications for copyright infringement claims against LLM companies. From a patent prosecution perspective, this study suggests that LLM companies may have understated the capabilities of their models, potentially leading to overbroad claims of non-infringement. Practitioners should be aware of this vulnerability when analyzing the patentability of LLM-related inventions or assessing the validity of existing patents. In terms of case law, this study may be relevant to the ongoing debate over the scope of copyright protection for AI-generated works (e.g., Google v. Oracle, 2021). The study's findings could be used to argue that LLMs, which rely on copyrighted works during training, may be infringing on those copyrights. Statutorily, this study touches on the issue of copyright infringement under Section 106 of the Copyright Act (17 U.S.C. § 106), which grants exclusive rights to the copyright owner. The study's findings could be used to argue that LLM

The regulatory approaches to open-source resources and software in China have significant implications for Intellectual Property (IP) practice globally. In comparison to the US, where the copyright laws governing open-source software and AI-generated products are primarily based on the fair use doctrine and the territorial principle of IP protection, China's approach to IP protection focuses on promoting an open-source development culture and addressing the challenges arising from AI system use. Internationally, the European Union and many countries follow a more nuanced approach, balancing the need for IP protection with the benefits of open-source collaboration and innovation. China's approach to IP protection in open-source software and AI-generated products, as outlined in the article, is distinct from the US and international approaches. While the US relies on the fair use doctrine to balance IP protection and open-source collaboration, China's approach prioritizes the promotion of open-source development culture and addressing the challenges arising from AI system use. This approach is also more aligned with the international trend of recognizing the importance of open-source collaboration and innovation in the development of AI-based solutions and services. However, the implications of China's approach on global IP practice and the potential for other countries, such as Russia and BRICS nations, to adopt similar models, require further analysis and consideration. In terms of jurisdictional comparison, the article highlights the following key differences: * **Copyright laws**: China's approach to copyright laws governing open-source software and AI-generated products is distinct from the US, where the fair use doctrine plays a significant role

As a Patent Prosecution & Infringement Expert, I analyze the article's implications for practitioners as follows: The article highlights the importance of considering China's regulatory approaches to open-source software and AI-generated products, particularly in the context of copyright protection and fair use. Practitioners should be aware that copyright laws governed by the territorial principle of IP protection determine the specific regimes for fair use of works, including computer programs. This is relevant to case law such as Sony Corp. of America v. Universal City Studios, Inc. (1984), which established the concept of fair use in the United States. The article also notes that China's approach to addressing key legal challenges arising from the widespread use of AI systems could serve as a reference for the development of AI legislation in other countries, such as Russia and the BRICS nations. Practitioners should be aware of statutory and regulatory connections, such as the Chinese Regulation on the Protection of Software Rights (2001), which provides a framework for protecting intellectual property rights in software, including open-source software. In terms of regulatory connections, the article implies that China's approach to open-source software and AI-generated products may be influenced by the country's national intellectual property strategy, as outlined in the 13th Five-Year Plan (2016-2020). This plan aims to promote the development of the digital economy and encourage innovation, which may have implications for patent prosecution and infringement strategies in the context of AI-generated products. Overall, the article suggests that

The article’s focus on AI-induced intersections with IP law resonates across jurisdictions, prompting nuanced responses. In the U.S., regulatory bodies like the USPTO have initiated public consultations on AI inventorship, aligning with evolving case law and market demands, reflecting a pragmatic, case-driven evolution. South Korea, by contrast, emphasizes proactive legislative frameworks, recently amending patent statutes to explicitly address AI-generated inventions, signaling a more codified, anticipatory approach. Internationally, WIPO’s ongoing dialogues on sui generis rights underscore a shared recognition of systemic gaps, yet implementation diverges: the U.S. favors flexible, interpretive adaptation, Korea leans toward statutory clarity, and Switzerland’s collaborative research model exemplifies a hybrid governance style—balancing academic insight with institutional policy development. These divergent trajectories illustrate how jurisdictional priorities—reactive versus codified, individual versus collective—shape the practical contours of AI-IP governance.

The article signals a pivotal shift in IP law as AI intersects with traditional rights, prompting practitioners to anticipate evolving jurisprudence on AI inventorship (patent law) and AI authorship (copyright law), particularly as case law begins to crystallize. Statutorily, these developments align with the need for sui generis protections under evolving IP frameworks, echoing precedents like *Alice Corp. v. CLS Bank* (2014) on defining patent eligibility for AI-driven innovations, while regulatory initiatives—such as policy projects by international bodies—may influence harmonization of sui generis regimes globally. Practitioners should monitor these intersections closely, as they may redefine IP ownership, attribution, and protection mechanisms.

The article on AI and copyright intersects with evolving jurisdictional frameworks, prompting nuanced analysis. In the US, copyright law traditionally requires human authorship, creating tension with AI-generated content; courts increasingly grapple with whether AI’s output constitutes original expression, as seen in cases like *Thaler v. Vidal*. South Korea, by contrast, has begun integrating AI into its copyright registry systems, permitting registration of AI-assisted works under specific attribution protocols, reflecting a pragmatic adaptation to technological advancement. Internationally, WIPO’s ongoing dialogue on authorship attribution—balancing human input with algorithmic generation—offers a harmonizing lens, though divergent national implementations reveal a spectrum from regulatory permissiveness (e.g., EU’s proposed AI-generated content directives) to constitutional constraints (e.g., U.S. First Amendment considerations). These comparative trajectories underscore that while the core challenge—authorship attribution—is universal, the legal responses reflect distinct cultural, economic, and institutional priorities.

The article highlights a critical intersection between AI and IPR, particularly in copyright, where AI’s capacity to generate creative works raises questions about authorship and ownership. Practitioners should anticipate increased litigation around AI-generated content, necessitating careful analysis of human intervention versus autonomous creation—a distinction that may hinge on precedents like the U.S. Copyright Office’s stance on human authorship or analogous jurisdictional frameworks. Statutorily, this aligns with evolving interpretations of “authorship” under copyright law, potentially influencing regulatory updates at WIPO or national agencies to accommodate AI’s role in innovation.

The article on Generative AI’s impact on copyright introduces a pivotal tension between traditional legal doctrines—specifically the idea-expression dichotomy and the substantial similarity test—and the emergent realities of AI-generated content. From a jurisdictional perspective, the U.S. approach tends to emphasize statutory interpretation and case-by-case adjudication, often deferring to precedent in determining ownership of AI-generated outputs, particularly when training data is sourced from public or licensed materials. In contrast, South Korea’s legal framework, influenced by its strong statutory codification under the Copyright Act, may more readily incorporate legislative amendments to address novel AI challenges, particularly concerning attribution and derivative rights, aligning with its proactive regulatory posture in digital innovation. Internationally, WIPO and EU discussions reflect a broader trend toward harmonizing definitions of authorship and originality in AI contexts, favoring a balance between protecting human creators and acknowledging the transformative role of AI as a tool rather than an author. These divergent yet converging trajectories underscore a global recalibration of IP norms, demanding adaptive legal frameworks that accommodate technological evolution without eroding foundational principles of authorship.

The article implicates practitioners by highlighting the tension between traditional copyright doctrines—specifically the **idea-expression dichotomy** and the **substantial similarity test**—and the emergence of AI-generated content. Practitioners must now contend with novel questions of **ownership attribution**, particularly regarding the **training dataset** used in generative AI models, which may redefine the scope of copyright protection under existing statutory frameworks (e.g., 17 U.S.C. § 102). Case law such as **Google LLC v. Oracle America, Inc.**, 141 S. Ct. 1183 (2021), may inform arguments on the permissibility of using pre-existing data for transformative outputs, while regulatory developments could emerge to address gaps in existing IP protections. Practitioners should anticipate litigation centered on dataset provenance and the applicability of copyright to AI-generated outputs.

The Stanford notification protocol aligns with U.S. DMCA requirements by designating a centralized agent for infringement claims, facilitating streamlined reporting consistent with 17 U.S.C. § 512(c). This mirrors international best practices seen in South Korea, where institutions similarly appoint designated agents under the Copyright Act’s Article 45 to coordinate infringement notifications, though Korea’s framework emphasizes broader statutory obligations on content hosts. Internationally, the trend reflects a convergence toward standardized reporting mechanisms—often codified in national statutes or institutional policies—to reduce administrative friction while preserving due process for rights holders. The Stanford model, while U.S.-centric, exemplifies a scalable template adaptable to diverse jurisdictional contexts without compromising procedural integrity.

The article establishes a clear procedural framework for reporting alleged copyright infringement at Stanford, aligning with DMCA requirements by designating an agent to receive notifications. Practitioners should note that compliance with statutory notice provisions—specifically including a good faith belief statement, accurate information under penalty of perjury, and identification of infringing material—is critical for validity and enforceability. This aligns with statutory mandates under 17 U.S.C. § 512 and reinforces case law precedent (e.g., *Perfect 10 v. Amazon*) that emphasizes procedural compliance for effective infringement claims.

The article’s impact on IP practice lies in its framing of SEP governance as a transferable model across emerging ecosystems—a critical insight given the convergence of automotive and telecom technologies. In the US, SEP licensing is largely governed by FRAND principles enforced through antitrust and contract law, fostering a relatively predictable dispute resolution framework. Korea, by contrast, integrates SEP oversight within its broader IP strategy, emphasizing proactive licensing agreements and administrative mediation, reflecting a more interventionist approach. Internationally, the WIPO and ITU provide harmonized guidelines that influence both jurisdictions, yet jurisdictional nuances persist, underscoring the need for adaptable legal architectures as V2X evolves. The article effectively bridges sector-specific precedent with systemic scalability, offering practitioners a template for navigating cross-industry IP conflicts.

The article draws a critical parallel between telecom SEP (Standard Essential Patent) battles and emerging V2X (Vehicle-to-Everything) ecosystems, highlighting the importance of early licensing strategies and standardization frameworks for IP practitioners. Practitioners should anticipate analogous disputes over essentiality and FRAND (Fair, Reasonable, And Non-Discriminatory) licensing obligations, akin to precedents like Huawei v. Interdigital (Fed. Cir. 2017), which clarified the enforceability of FRAND commitments. Statutorily, this aligns with 35 U.S.C. § 271(e) implications for standardization and compulsory licensing in technology ecosystems. The V2X context demands proactive IP planning to mitigate litigation risks similar to those in telecom.

**Jurisdictional Comparison: CRISPR Gene Editing Technology Patents** The ongoing patent disputes surrounding CRISPR gene editing technology have sparked a global debate on intellectual property strategy, innovation, and patient access. In this context, a comparison of US, Korean, and international approaches reveals distinct differences in patent law and regulatory frameworks. **US Approach:** The US Patent Trial and Appeal Board's (PTAB) ruling in favor of the Broad Institute for eukaryotic applications reflects a more permissive approach to patent claims, allowing for broader protection of CRISPR-Cas9 technology. In contrast, the Federal Circuit's decision in _Board of Regents of the University of Wisconsin System v. Synopsys, Inc._ (2020) underscored the importance of written descriptions in patent claims, potentially limiting the scope of CRISPR patents. **Korean Approach:** South Korea's patent law has taken a more nuanced approach, recognizing the importance of CRISPR technology while also addressing concerns over regulatory frameworks and public access. The Korean government has established a regulatory framework for gene-edited crops, which may influence the development of CRISPR-based therapies in the country. **International Approach:** The European Patent Office's (EPO) varying positions on related patents reflect a more restrictive approach to patent claims, emphasizing the need for clear and concise descriptions of inventions. In contrast, the World Intellectual Property Organization (WIPO) has taken a more neutral stance, promoting international cooperation and harmonization of patent

As a Patent Prosecution & Infringement Expert, I will provide a domain-specific expert analysis of the article's implications for practitioners. The ongoing patent disputes surrounding CRISPR gene editing technology have significant implications for biotech innovation, patient access, and IP strategy. The article highlights the complexity of the patent landscape, with multiple institutions claiming overlapping rights in next-generation editing tools, therapeutic applications, and agricultural applications. This complexity is likely to lead to an increase in patent litigation and disputes, which can be mitigated with effective IP strategy and licensing approaches. **Case Law Connection:** The U.S. Patent Trial and Appeal Board's (PTAB) ruling in favor of the Broad Institute for eukaryotic applications may be seen as analogous to the Supreme Court's decision in **Alice Corp. v. CLS Bank International** (2014), which emphasized the importance of determining the patentability of abstract ideas. However, the PTAB's decision may also be contrasted with the **Myriad Genetics** case (2013), where the Supreme Court held that isolated DNA is not patentable subject matter. **Statutory Connection:** The article highlights the importance of understanding the patent landscape in the biotechnology industry, which is governed by the **Patent Act of 1952** (35 U.S.C. § 101 et seq.). The statute provides the framework for determining patentability, including the requirement that a patent must claim a "new and useful process, machine, manufacture, or composition

The article’s impact on Intellectual Property practice is nuanced across jurisdictions, reflecting divergent regulatory philosophies. In the U.S., the Copyright Office’s stance on AI as an ineligible author—rooted in statutory interpretation of “authorship” under 17 U.S.C.—creates a clear boundary, yet leaves room for litigation over human-AI collaborative outputs. South Korea, by contrast, leans toward a functionalist approach, permitting registration of AI-generated works where human oversight is demonstrable, aligning with broader East Asian regulatory pragmatism. Internationally, WIPO’s ongoing dialogues underscore a consensus-building trajectory toward recognizing “creative contribution” as a threshold for attribution, balancing innovation incentives with authorial accountability. These divergent paths—U.S. textualism, Korean contextualism, and global harmonization—highlight the evolving imperative for legislative adaptability without compromising core copyright principles.

The article implicates practitioners to reassess copyright frameworks in light of AI’s expanding role in content creation, particularly regarding authorship attribution and legal responsibility for AI-generated works. Practitioners should consider precedents like *Naruto v. Slater* (2018), which addressed non-human authorship, and statutory considerations under copyright regimes that define authorship eligibility—issues now contested in AI contexts. Regulatory adaptation, as highlighted, necessitates aligning legislative definitions of “authorship” with evolving AI capabilities to balance protection and innovation.

The Minnesota Law Review articles offer nuanced intersections with Intellectual Property, particularly through the lens of doctrinal alignment and separation of powers. Tun-Jen Chiang’s critique of the incentive-to-invent theory in trade secret law resonates across jurisdictions, as similar debates arise in the U.S. and Korea regarding the balance between incentivizing innovation and protecting proprietary rights—though Korea’s statutory framework leans more explicitly toward statutory exclusions, while the U.S. emphasizes equitable doctrines. Meanwhile, Hoffman and Swedloff’s analysis of consumer contract boilerplate, though not IP-specific, informs IP practice by highlighting how procedural constraints (e.g., arbitration clauses) may similarly limit enforceability of IP-related dispute resolution clauses, a concern echoed in international arbitration forums. Internationally, the trend toward harmonizing IP protection through WIPO and TRIPS frameworks aligns with the doctrinal critiques in these pieces, suggesting a shared imperative to align procedural fairness with substantive rights. Thus, the symposium’s broader themes—separation of powers, doctrinal coherence, and procedural equity—offer cross-jurisdictional relevance to IP practitioners navigating both domestic and global legal landscapes.

The article’s implications for IP practitioners hinge on intersecting principles of statutory interpretation and separation of powers. The invocation of the major questions doctrine in administrative law, as discussed by Johnson, parallels its potential application in IP contexts where courts review agency interpretations of patent statutes—reinforcing the judiciary’s role in safeguarding constitutional boundaries. Similarly, Chiang’s critique of the incentive-to-invent theory resonates with IP jurisprudence that increasingly scrutinizes statutory structures of trade secrets and patents for alignment with legislative intent, as seen in cases like *Kewanee Oil Co. v. Bicron Corp.* (1974), which emphasized statutory purpose over economic theory. These threads suggest a broader trend: courts are more willing to interrogate statutory frameworks through separation-of-powers lenses, impacting both administrative and IP litigation. Practitioners should anticipate heightened scrutiny of statutory authority and doctrinal consistency in both patent and trade secret cases.

The article’s focus on watermarking as a mechanism to protect deep neural networks’ intellectual property resonates across jurisdictions, yet implementation nuances diverge. In the U.S., where copyright law extends to original compilations of data and software, watermarking may be recognized as a supplementary layer of protection, potentially qualifying for statutory damages if infringement is proven—though enforcement remains contingent on proving ownership and unauthorized reproduction. In South Korea, where intellectual property protection is robust and increasingly aligned with digital innovation, courts have begun to acknowledge embedded identifiers as indicators of authorship in software-related disputes, offering a more receptive legal framework for watermark-based claims. Internationally, WIPO and EU directives emphasize the need for technical safeguards that preserve anonymity while enabling verifiable attribution, aligning with the article’s premise but urging broader standardization of watermarking protocols to avoid fragmentation of legal recognition. Thus, while the conceptual utility of watermarking is universally acknowledged, jurisdictional divergence in recognition of embedded identifiers as legally actionable evidence presents a critical challenge for harmonized IP protection in AI-driven contexts.

This article implicates practitioners by highlighting the growing need for IP protection mechanisms tailored to deep neural networks, a critical gap in current AI IP frameworks. Watermarking as a solution aligns with statutory protections under copyright law (e.g., 17 U.S.C. § 102) for original works of authorship, potentially extending applicability to AI models as “works” under existing legal definitions. Practitioners should monitor case law developments, such as those analogous to software copyright precedents (e.g., Oracle v. Google), to anticipate how courts may treat embedded watermarks as evidence of ownership or infringement. The regulatory implication is clear: IP protection strategies for AI must evolve alongside technological innovation to remain enforceable.

The Minnesota Law Review's subscription and copyright policies have implications for Intellectual Property (IP) practice, particularly in the context of academic publishing. In comparison to the US approach, Korean copyright law tends to favor stricter controls on copyright permissions, whereas international approaches, such as the Berne Convention, emphasize the importance of balancing authors' rights with public access to information. The Minnesota Law Review's policy of allowing duplication of articles for classroom use, provided the author has not retained copyright, reflects a more permissive approach, similar to US fair use provisions, whereas Korean law might require more explicit permission from the copyright holder.

Analysis: This article appears to be a standard copyright notice and subscription policy for the Minnesota Law Review. However, from a patent prosecution and infringement expert's perspective, there are no direct implications for patent law. Nevertheless, it can be noted that the article's copyright notice and subscription policy might be relevant in the context of fair use provisions under 17 U.S.C. § 107, which could potentially impact patent-related publications or research. Implications for Practitioners: This article does not have any direct implications for patent practitioners, but it highlights the importance of understanding copyright and subscription policies when utilizing or referencing copyrighted materials, including patent-related publications. Patent practitioners should be aware of fair use provisions and copyright laws when using copyrighted materials in their work. Case Law Connection: This article does not have any direct case law connections, but it is relevant to the broader context of copyright law, which intersects with patent law in areas such as patent-related publications and research.

**Jurisdictional Comparison and Analytical Commentary** The proposed Model Locking (M-LOCK) scheme for deep neural networks (DNNs) enhances the availability protection of learning models, a crucial aspect of intellectual property (IP) protection in AI-based cybersecurity services. A comparison of the US, Korean, and international approaches to IP protection in AI reveals distinct differences in their treatment of IP rights in machine learning models. In the **United States**, the current IP framework does not directly address the protection of AI-generated works, including machine learning models. The US Copyright Act of 1976 protects original works of authorship, but the application of this framework to AI-generated works is still unclear. The proposed M-LOCK scheme may be seen as an innovative solution to this issue, providing a means of protecting IP rights in AI-generated works. However, the US courts may need to interpret and apply existing laws to determine the validity of such protection. In **Korea**, the IP protection of AI-generated works is more developed. The Korean Copyright Act (2011) protects AI-generated works as "computer-generated works," providing a clear framework for their protection. The proposed M-LOCK scheme may be seen as a complementary measure to enhance the availability protection of learning models, which is not explicitly addressed in the Korean Copyright Act. Internationally, the **Berne Convention for the Protection of Literary and Artistic Works** (1886) and the **Agreement on Trade-Related Aspects of Intellectual Property

As a Patent Prosecution & Infringement Expert, I'll provide domain-specific expert analysis of the article's implications for practitioners. The article discusses the need for intellectual property protection in AI-based cybersecurity services, particularly in deep neural networks (DNNs). The proposed "Model Locking" (M-LOCK) scheme and "Data Poisoning-based Model Manipulation" (DPMM) method aim to enhance availability protection and prevent piracy. This is relevant to patent practitioners as it highlights the importance of protecting intellectual property in AI-based technologies, which may involve novel and complex methods for protecting DNNs. From a patent perspective, the M-LOCK and DPMM schemes may be considered as novel methods for protecting intellectual property in AI-based systems. Patent practitioners may need to consider how to claim and protect these methods, potentially involving novel combinations of machine learning and security techniques. Relevant case law, such as the Supreme Court's decision in Alice Corp. v. CLS Bank Int'l, 573 U.S. 208 (2014), may be relevant in determining the patentability of these methods. In terms of statutory and regulatory connections, the article touches on the importance of protecting intellectual property in AI-based systems, which is a key aspect of the Leahy-Smith America Invents Act (AIA) and the Patent Act of 2011. The proposed M-LOCK and DPMM schemes may also be relevant to the development of new standards for AI-based systems, which may be influenced

The concept of authorship in AI-generated works poses significant challenges to copyright law, with jurisdictions such as the US and Korea adopting distinct approaches to determining originality and ownership. In contrast to the US, which tends to focus on human authorship, Korea has shown a more nuanced stance, considering the potential for AI systems to be deemed co-authors, whereas international approaches, such as those outlined in the Berne Convention, emphasize the importance of human creativity and originality. Ultimately, the lack of uniformity in addressing AI-generated works underscores the need for a harmonized global framework to clarify the complexities surrounding text prompts, originality, and copyright protection.

The article's exploration of originality in AI-generated works and text prompts has significant implications for copyright law practitioners, particularly in light of cases such as Aalmuhammed v. Lee (1999) and Feist Publications v. Rural Telephone Service (1991), which established the importance of human authorship and originality in copyright protection. The article's analysis of philosophical foundations of copyright and collage protection may also inform discussions around the Copyright Act of 1976 and the Digital Millennium Copyright Act (DMCA), which govern copyright law in the United States. Furthermore, the article's focus on the correlation between text prompts and resulting outputs may raise questions about the applicability of Section 102(a) of the Copyright Act, which requires that a work be "fixed in any tangible medium of expression" to be eligible for copyright protection.

The article highlights the growing intersection of technology, policy, and intellectual property (IP) law, prompting a comparative analysis of US, Korean, and international approaches. In the US, the ongoing struggle over tech platform policies, as seen in the example of Apple's App Store and Meta's Oversight Board, reflects the evolving landscape of IP regulation. The US approach tends to emphasize a balance between free speech and IP protection, with government regulators and civil society groups playing significant roles in shaping policies. In contrast, South Korea has taken a more proactive stance on regulating tech giants, exemplified by the recent amendment to the Telecommunications Business Act, which empowers the government to regulate and impose penalties on tech companies that fail to comply with data protection and IP laws. This approach reflects a more interventionist regulatory framework, where the government plays a more active role in shaping tech policies. Internationally, the European Union's Digital Services Act (DSA) and the General Data Protection Regulation (GDPR) have set a precedent for regulating tech platforms, emphasizing the need for transparency, accountability, and IP protection. The DSA, in particular, requires tech platforms to take more proactive measures to prevent the spread of misinformation and protect users' rights, including their IP rights. These jurisdictional differences highlight the complexities of IP regulation in the digital age, underscoring the need for a nuanced and context-specific approach that balances competing interests and values.

As a Patent Prosecution & Infringement Expert, I will analyze the article's implications for practitioners in the context of intellectual property law. The article highlights the intersection of technology, policy, and intellectual property, particularly in the realm of AI and content moderation. This intersection is relevant to patent practitioners in the following ways: 1. **Copyright and Patent Infringement**: The article mentions Disney and Paramount alleging that ByteDance's AI model, Seedance 2.0, is distributing and reproducing their intellectual property. This raises questions about the liability of AI systems for copyright infringement and patent infringement. Practitioners should be aware of the potential for AI-driven infringement and the need for companies to develop robust content moderation policies to mitigate these risks. 2. **Patent and Technology Policy**: The article also touches on the intersection of patent law and technology policy, particularly in the context of AI and national security. The potential designation of Anthropic as a "supply chain risk" by the Department of Defense raises questions about the impact of national security regulations on patent law and the development of AI technologies. Practitioners should be aware of the potential for government regulations to influence patent law and the need for companies to navigate these complex regulatory landscapes. 3. **Patent Prosecution and AI**: The article highlights the rapid development of AI technologies and the need for companies to develop robust patent strategies to protect their innovations. Practitioners should be aware of the potential for AI-driven inventions and the

The advent of large language models (LLMs) has introduced a novel vulnerability in content-aware semantic watermarking schemes, which are commonly used to protect intellectual property in online platforms. This vulnerability, exposed by the Coherence-Preserving Semantic Injection (CSI) attack, highlights the need for updated security measures in the US, Korean, and international approaches to intellectual property protection. In the US, the Digital Millennium Copyright Act (DMCA) and the Copyright Act of 1976 provide a framework for protecting intellectual property, but may not adequately address the CSI attack. In contrast, Korean law, such as the Copyright Act of 2019, may provide more comprehensive protection for semantic watermarking, but its applicability to LLM-driven attacks is uncertain. Internationally, the Berne Convention and the WIPO Copyright Treaty may provide a basis for protecting intellectual property, but their implementation and enforcement vary across jurisdictions. The CSI attack's success in invalidating content-aware semantic watermarking bindings highlights the need for more robust security measures, such as the use of multi-modal watermarking or the integration of AI-powered detection systems. This development has significant implications for intellectual property practitioners, who must now consider the potential for LLM-driven attacks when advising clients on content protection strategies. Moreover, the CSI attack's reliance on LLMs raises questions about the role of AI in intellectual property protection and the need for updated laws and regulations to address these emerging technologies. In terms of jurisdictional comparison, the US, Korean, and

As a Patent Prosecution & Infringement Expert, I'll provide domain-specific expert analysis of this article's implications for practitioners. **Analysis:** The article discusses a novel attack, Coherence-Preserving Semantic Injection (CSI), that exploits the vulnerability of content-aware semantic watermarking schemes in generative images. The CSI attack leverages large language models (LLMs) to manipulate semantic spaces, enabling targeted and coherent alterations that invalidate watermark bindings. This attack is particularly effective against current semantic watermark designs, which rely on binding watermark signals to high-level image semantics. **Implications for Practitioners:** 1. **Patentability of semantic watermarking techniques:** The article highlights the limitations of current semantic watermarking schemes, which may impact their patentability. Practitioners should carefully consider the prior art and existing patents when developing and prosecuting patent applications related to semantic watermarking techniques. 2. **Security and validity of existing patents:** The CSI attack raises concerns about the security and validity of existing patents that rely on content-aware semantic watermarking. Practitioners should review their existing patent portfolios and consider potential vulnerabilities to CSI attacks. 3. **New patent applications:** The article's findings may inspire new patent applications that address the vulnerabilities of current semantic watermarking schemes. Practitioners should consider developing and prosecuting patent applications that incorporate novel techniques for secure and robust semantic watermarking. **Case Law, Statutory, or Regulatory Connections:** 1. **35 U.S.C. § 101