When AI Adoption Outpaces IP Protection: The Risks Companies Must Address | IPWatchdog Unleashed

This week on IPWatchdog Unleashed, we have a discussion about the IP risks associated with AI use and adoption, with Stephanie Curcio, Marlene Valderrama, Kendra Ervin and Rinat Shangeeta. Recorded at our recent Women’s IP Forum, the episode features my introductory remarks followed by a panel led by Curcio . The discussion reveals a consequential gap with how businesses are deploying AI. Many companies are moving faster to adopt AI than their governance, contracting and IP review processes can accommodate. For those responsible for protecting innovation and confidential information, the challenge is ensuring that enthusiasm for AI adoption does not obscure the consequences of how these tools are actually used.

One immediate concern is the growing use of AI to generate invention disclosures and other substantive work product. A lengthy, polished disclosure does not necessarily make the underlying invention easier to identify or evaluate. In fact AI can sometimes obscure what the inventor actually contributed. The panel discusses the importance of adapting intake procedures to understand where AI was used and distinguish human contributions from AI generated material. For patent practitioners and R&D leaders, that distinction matters to inventorship analysis and to understanding precisely what the organization should seek to protect.

AI adoption also requires budget discipline. The panel discusses “token maxing,” which describes using large volumes of AI processing to demonstrate activity, even when that consumption does not translate into useful work. As organizations confront the resulting bills, the emphasis is shifting toward matching the model to the task, establishing usage limits, and tailoring access to different roles. The practical lesson is that AI consumption is not a measure of productivity. Spending should support defined business and legal objectives, with professional judgment determining whether the output justifies the cost.

Confidentiality also presents another immediate challenge, particularly when employees use unauthorized tools or enter sensitive information into systems without understanding their terms. This so-called shadow AI problem reaches beyond public chatbots to specialized tools, vendor trials and unapproved uses of otherwise authorized platforms. The discussion emphasizes that approving a tool does not resolve every question about permissible data or use cases. For law firms and service providers, client restrictions add another layer. And regardless of internal authorization, use allowed within a company or firm cannot substitute for compliance with obligations governing a client’s information.

This means effective governance requires more than a written policy. In this regard, the conversation explores mapping AI use across workflows, providing accessible approved alternatives, and educating employees about both technical capabilities and legal boundaries. Agentic AI makes access controls especially consequential because these systems can interact with tools and information beyond an individual prompt. For organizations handling sensitive engineering, software and research data, permissions must reflect what a particular task requires. IP professionals need a place in those governance discussions before deployment decisions are finalized.

Vendor contracting is equally important. Rights to use inputs for training, ownership or control of outputs, confidentiality protections, and infringement indemnification all affect the organization’s exposure. The panel cautions that indemnification provisions require scrutiny: exclusions and conditions may leave the intended use case without meaningful coverage. Those questions can arise during evaluation, before a company purchases a subscription. A trial involving actual client or proprietary information deserves legal and security attention before that information enters the system.

The conversation also addresses the human judgment needed to make AI useful. Generated contract revisions and expansive risk assessments still require someone who understands the transaction, the business and the relevant legal priorities. More output does not necessarily mean better work. Organizations must also address employee concerns about displacement while explaining why professional expertise remains essential to evaluating results. Adoption becomes more productive when the right people use appropriate tools for defined tasks and remain accountable for the work.

For in-house counsel, patent practitioners and technology leaders, the central takeaway is that IP protection must be part of AI implementation from the outset. Decisions about access, data, vendors and workflows can have consequences long before a dispute arises. This episode offers practical perspectives on closing that gap while preserving the benefits that make AI adoption compelling.

More IPWatchdog Unleashed

You can listen to the entire podcast episode by downloading it wherever you normally access podcasts or by visiting IPWatchdog Unleashed on Buzzsprout. You can also listen to IPWatchdog Unleashed conversations on the IPWatchdog YouTube channel. For more IPWatchdog Unleashed, see below for our growing archive of previous episodes.