IP and the Wild West Landscape of AI | IPWatchdog Unleashed

This week on IPWatchdog Unleashed I speak with Allison Gaul who serves as legal counsel for Boston Consulting Group. She is responsible for evaluating digital products with an eye towards intellectual property strategy, value creation, and legal risk. She is also a recovering, or at least former patent attorney. She does still advise BCG on patent issues, but she is not drafting and prosecuting patent applications at this point. We begin our conversation with me asking about what she believes are the biggest legal issues in the IP world today.

“This is such a hard question to tackle,” Gaul said pensively. “I feel like right now when we’re in this wild west kind of landscape of AI developing so quickly and Web 3.0 popping up beside it and quantum coming up hard behind, it’s like we’re in some kind of AI horse race. And then this regulatory patchwork that’s just blooming all over the globe from both an AI perspective and a data perspective. I think at any moment in time, I feel like I’m playing whack-a-mole. Like, what is today’s hot issue.”

Gaul did identify several things that stay top of mind for her in terms of IP rights at the intersection of AI, with various issues relating to data front and center as the top issue.

“One is data control. Input data, that’s rights and integrity,” Gaul explained. Output data, rights and integrity, security, privacy, and then also jurisdictional components because we’re seeing more and more jurisdictions start to have restrictions on what can come in and out of their jurisdiction and what can be stored there, what needs to have some kind of validation by the government before it can actually be moved in and out of the country.”

The second area identified by Gaul was open source, which she explained doesn’t usually have a lot of changes because it is normally “a slow-moving dinosaur”, but recently “we’ve really seen a proliferation of companies that are putting out AI models as open source, when what they’re really doing is kind of putting out a version of it. Maybe it’s the model, but not the weights. Maybe it’s the weights, but not the model…The training data isn’t available. So, it’s not necessarily true open source… And that’s not necessarily a new concept, dual licensing. We’ve seen Red Hat and we’ve seen Oracle do this really well in other companies over the years. But there’s something a little bit insidious maybe about the way that it’s happening now in this sort of AI arms race and rush to market share, where people are putting stuff out there just to try to gobble up as much user base as they can, and then doing a little bit of a switcheroo.”

The third and final thing that Gaul identifies as being constantly top of mind is the overall speed of AI development.

Returning to the beginning of her list, I seed the conversation by acknowledging the privacy concerns that have grown over the last decade, highlighted most prominently by the Europeans. And now with a company like 23andMe going bankrupt there are concerns about what will become of all the data they have been giving and have accumulated.

“I think the ownership of data and the use rights for data are actually one of the paramount questions in innovation right now,” Gaul said. “[E]very innovation needs data, like whether it’s quantum applications, whether it’s metaverse, whether it’s web 3.0, whether it’s different types of AI, not just generative, everybody needs data.”

“We were seeing for a while this real escalation towards better and better privacy protections around the world in different jurisdictions,” Gaul explained. “But now this year, we’re actually seeing the EU saying they’re going to start simplifying GDPR, some other jurisdictions saying they’re going to start backing off of some of it a little bit because they feel it’s stifling innovation within the AI race.”

The conversation continued over privacy, informed consent and whether users actually read terms and conditions and know what they are agreeing to.

“Right now, I think we’re just in this really awkward space where everything’s kind of growing in different ways, like a rapidly growing like tree,” Gaul said. “But I think what’s going to happen is that as these larger AI companies start to come under more and more legal scrutiny and more and more fire, and people get more and more upset about it online, about the way that their data is being treated, or the way that they’re not able to understand transparently how these things work, is that we’re going to start to see people relying more and more actually on smaller tools. And so, there’s just hundreds and hundreds and hundreds of small AI companies popping up.”

I agree with Gaul’s assessment, and I think we’re going to have a handful of very large companies that provide the backbone of future AI tools. These large AI giants will be very good at machine learning and very good at digesting massive amounts of information. Then we will likely see silos of expertise established. These silos or niches will be dominated by small companies that operate within a niche industry that they know really well. Indeed, we are already seeing small companies developing specific tools within the IP space that we are quite familiar with, which are much better for a specialized purpose because they understand the IP industry and what the industry wants and needs. Based on what we know is happing in the IP sector it seems entirely likely that smaller companies across many industries will find success focusing on niche applications specific to particular industry and deliver on a subset relevant needs.

“As we move in that direction, I do think that’s going to solve some of our IP problems, because it’s going to make it a lot easier to understand the implications of licensing a piece of IP for use with a particular app or program or model,” Gaul explained. “I think that’s going to alleviate some of the challenges that we’re having, not necessarily around the protectability of these things, but how we at least manage licensing and attribution for different types of assets when we have these specific tools that are using them in specific ways.”

We go on to discuss fair use, particularly discussing the legal troubles facing Meta, and ethics around AI development and use, as well as the importance of prompts and how it is frustrating—to say the least—that AI companies do not seem interested in helping users learn how to get better at prompting AI tools to get better results.

And on that happy note our podcast ends and the live panel transitioned into taking more questions from our studio audience. You can listen the entire podcast episode by downloading the podcast wherever you normally access podcasts, by visiting IPWatchdog Unleashed on Buzzsprout or by watching the conversation on the IPWatchdog YouTube channel.

More IPWatchdog Unleashed

For more IPWatchdog Unleashed, see below for our growing archive of previous episodes.