Automating Extortion: How Advancements in AI Technology Could Supercharge Abuse of IP Rights
Since the unveiling of consumer-facing generative tools, artificial intelligence has dominated discussion in lecture halls, board rooms, and courtrooms across the globe, with implications in seemingly every industry. This has, of course, been no exception in the US; though, while we’ve all seen countless headlines, reports, and proposals for legislation highlighting the potential risks of AI, little substantive law has actually been adopted at the federal level to safeguard against abuse, despite a now multi-year maturation period.
It is certainly not uncommon for the law to lag behind emerging technologies—in fact, it typically does. Some commentators even view this natural evolutionary gap between the two as being beneficial. After all, lawmakers consequently have more time to properly assess the implications of new innovations, thereby reducing the possibility of unforeseen ramifications resulting from hastily concocted policy. While there is absolutely merit to this line of thinking, it is becoming increasingly obvious that with AI, however, we may no longer have the luxury of time given its rapid advancement.
This is particularly true with respect to AI in the field of intellectual property, where recent technological developments seem increasingly likely to fuel fires that have long been smoldering on the back-burners of IP discourse. Specifically, AI is poised to greatly facilitate abuse of the respective intellectual property legal frameworks, supercharging patent trolling, copyright trolling, trademark bullying, and other predatory behaviors—many of which commentators believe are underregulated to begin with.
Ultimately, the likely outcome will be a stifling of innovation and creativity as bad-actors use AI tools to systematically target and extort vulnerable low-resource entities—leaving everyone from individual artists and mom-and-pop shops to promising tech startups in the crosshairs.
Thus, it is crucial that meaningful safeguards are quickly put in place in the intellectual property legal frameworks to protect grassroots innovation.
What’s Changed?
In short, recent advancements in AI’s ability to reason, remember, and interact have helped it break out of the confines of the chat box and venture into the real world. While some of these advancements consist of improvements to existing processes—making AI more efficient and precise in general—others introduce entirely new use cases that shift AI technologies from passive tools used for writing funny poems about your pets to a mechanism for the creation of autonomous industrial-scale webs of goal-oriented virtual workers.
Improvement on advanced reasoning methods — Tree-of-thoughts reasoning
Tree-of-thoughts reasoning, introduced in 2023, expands upon previous reasoning methods like chain-of-thoughts by exploring multiple logical paths in a branching structure, allowing models to evaluate and refine a number of solutions before selecting the best one. This has greatly improved problem-solving ability. Additionally, improvements in reinforcement learning technology help models continuously refine these improved decision-making strategies in specific environments by tracking patterns in success rates.
Long-term memory and context retention improvements
First seen meaningfully in Claude AI in 2023, these improvements allow AI models to retain and recall relevant information across extended interactions, enabling evidence-based reasoning, better adaptation to user needs, and continuity in complex workflows. Models can now retain past actions, learn from environmental feedback, and work in context, whereas older AI models were much more prone to “forgetting” previous interactions.
Function calling and API integration
Function calling enables AI models to seamlessly integrate into external programs, allowing them to execute precise, goal-oriented tasks while leveraging AI’s creative capabilities where beneficial and constraining them where necessary. This mechanism also facilitates direct interaction with external APIs and databases, such as the USPTO’s OpenData Portal or Thomson Reuters’ Dockets API, enabling AI-driven systems to retrieve, analyze, and act on specialized real-world information in real time.
Finally, though there are more, Multimodal AI technology extends AI’s recognition capabilities beyond simple text, enabling it to analyze, process, and integrate multiple datatypes like images, videos, audio, and complex visual patterns. This allows models to identify patterns and extract insights from a much larger pool of information, even when it might be altered, obscured, or restructured.
The Result: AI Agents and Multi-Agent Systems
In conjunction, these advancements have led to the creation of “AI agents” that are able to analyze, make decisions, and act across multiple domains. These agents can then further be combined into supercharged multi-agent systems, operating somewhat similarly to a business with multiple departments. Developed in frameworks like CrewAI, these weblike systems allow specialized AI agents to handle different aspects of complex workflows.
When considering the intellectual property law, It’s not hard to imagine how armies of AI agents might be used for extortion. What once required manual effort—scouring the internet for potential infringers, drafting demand letters, issuing takedown notices, etc.—can now all be taken care of while a system’s human creator is quite literally fast asleep. Virtually every step of an IP extortion scheme could, in theory, be completely automated.
For example, in a multi-agent copyright trolling system, one agent could continuously scan the internet or specific databases for potential violations with unprecedented efficiency—identifying images, text, or audio that vaguely resembles protected content. Another agent could use that information to assess legal arguments and gather compromising data on specifically selected vulnerable targets, drafting personalized demand letters tailored to maximize pressure. Finally, a third could monitor responses, refine negotiation strategies based on those responses, and apply specific tactics based on past interactions. Meanwhile, reinforcement learning algorithms will be optimizing these efforts in real time, continuously improving accuracy and efficiency for the next victim.
An Unfair Matchup
Amidst the advancement in technology, one thing remains unchanged, however—small businesses, independent creators, and local entrepreneurs will continue operating in much the same way they always have, often with limited budgets and little access to legal expertise. The local ice cream shop crafting homemade flavors, the quilting studio pouring time and care into every stitch, or the landscaping business that thrives on word-of-mouth referrals—all of them exist far outside the high-tech arms race shaping the modern intellectual property battlefield. These businesses are focused on serving their communities, not navigating the complexities of AI-driven legal threats. Yet, they are the ones most vulnerable to exploitation. They won’t have access to sophisticated AI-driven legal tools or expansive legal teams to fend off automated waves of cease-and-desist letters.
That is why it is imperative that safeguards be put in place—before these tactics become too deeply entrenched. Proactive legal protections, increased transparency in automated enforcement systems, accessible resources, and cost-effective dispute resolution mechanisms for small businesses must be prioritized to level the playing field. Without these measures, the rapid progress of AI will actively put low-resource entities at risk, turning a technology meant to foster innovation into tools for innovation-stifling abuse.
Graham Davis
Graham is currently a 2L at UNC Law with a strong interest in the intersection of emerging technologies and intellectual property law. In addition to being a staff member on the North Carolina Journal of Law and Technology, Graham also serves as an executive board member for the Carolina Intellectual Property Law Association (CIPLA).