Don’t miss out on news you won’t find anywhere else — get The Memo US and The Memo Europe in your inbox every two weeks.

When Mattel and OpenAI unveiled a high-profile collaboration in June allowing for the toy maker to use AI tools to develop new products from its existing brands, it sparked a variety of reactions ranging from excitement over the creative possibilities to concerns about potential security risks.

Those reactions capture the mixed mood in leveraged finance over the AI boom. The influx of AI technologies — or at least its current form — has largely benefitted LevFin bankers, who have been busy helping borrowers like DuPont spin-off Qnity Electronics and Invenergy’s Lackawanna site in Pennsylvania syndicate debt to fuel the skyrocketing demand for AI infrastructure.

That same demand has also given distressed companies like Altice USA the opportunity to raise cash through securitizing its hybrid-fiber coaxial assets (an important building block for AI infrastructure) without resorting to liability management exercises, as 9fin reported last week.

But for debt investors, the increasing number of leveraged credits partnering up with AI firms has also raised concerns about credit risks. One important discussion that’s taking shape is how AI-generated products are treated as potential debt collateral.

That issue is especially relevant for Mattel, which will soon need to address roughly $600m of SUNs coming due in April 2026. Intellectual property — one of the assets most exposed to AI disruption — is central to the company’s business model. As Mattel continues to push its identity as an “IP-driven company” (according to its website) and embarks on its OpenAI partnership, questions around debt underwriting and the pricing of AI-related risks are likely to come to the forefront.

Partnerships matter

Mattel’s tie-up with OpenAI may be high profile, but it isn’t the first in the LevFin universe. Other leveraged issuers in the media and entertainment space have also joined forces with AI businesses to refine existing content and speed up production, including:

• Lionsgate and AI research company Runway announced a partnership last September where Runway will create and train a new AI model “customized” to Lionsgate’s proprietary portfolio of film and TV content to aid pre and post-production processes
• iHeartMedia and enterprise AI platform Veritone; since February 2022, iHeart has used Veritone’s synthetic voice technology to translate and produce podcasts for new markets
• Veritone also partnered with Creative Artists Agency last May to launch “TheCAAVault” — a “secure repository” that stores the intellectual property of artists’ names, images and likenesses, including digital scans and voice recordings.

These companies haven’t disclosed much detail beyond the initial announcements. When approached by 9fin, most of these firms, including Mattel and OpenAI, did not respond to requests for comment except for CAA which declined to comment. Mattel’s June press release did indicate that the first product from the OpenAI collaboration is expected to be announced “later this year”.

But as these partnerships become more commonplace, one key question emerges — who gets legal ownership of the AI-generated assets?

For Mattel at least, published reports indicate that the toy maker will not be licensing its IP to OpenAI, and it will remain in control of the products being created.

A key rule in copyright law remains — AI-generated content can be copyrighted if there is proof of substantial human input, as pointed out by Sumeet Gupta, senior managing director and leader of AI and digital transformation at FTI Consulting.

“If you have a simple document that’s written mostly by AI, that will likely not fall within copyright claims,” Gupta said in a conversation with 9fin. “But if you have text that is partly written by AI but then sufficiently supplemented or transformed by human written text, and you have some auditable record of how that was created, the human part can muster through the copyright claims.”

Companies also need to choose their AI partnerships carefully as that could also affect their risk assessment when they raise debt in the future.

“If you have AI software that’s being hosted in a place that’s a geographically high-risk area, then it may have an impact based on the risks associated with data leakage and cyber security in that area,” said Marguerite McConihe, patent strategist and litigator at Mintz. “If you’re using a platform that is scraping and providing data that can’t be sourced, you’re also creating risk for yourself.”

Evolution or revolution?

Perhaps the most exciting part of AI is its ability to stretch human creativity — but can underwriters embrace AI-generated products as a viable form of debt collateral?

As we previously noted, IP is one of the assets that could be significantly disrupted by AI. That poses problems for private equity firms and companies which have spent billions of dollars acquiring and/or securitizing music IP collateral.

Warner Music Group, which has one of the largest music catalogs in the industry, is a case in point. When the company repriced its $1.295bn TLB due 2031 last September, Fitch Ratings flagged generative AI as a key rating driver in its ratings report: “AI can enhance operational efficiencies and artist development, but there are also potential risks from unlicensed AI training on copyrighted works and deepfakes from AI-generated music,” the report stated. WMG did not respond to a request for comment.

As it stands, using AI tools in music production is still perceived as controversial, even when it’s not being used to create deepfakes — just witness the backlash from fans when The Beatles released an AI-enhanced single created from demo tapes back in 2023.

But aside from production, there are other ways in which AI is already shaping how the music industry operates and is financed. Investors are making use of AI tools to streamline data collection processes that shortens the securitization timeline for existing pools of music IP, for example.

“The rise of AI has enhanced transparency and valuation accuracy by enabling real-time analysis of streaming data, listener demographics and engagement trends,” said Mark Boidman, partner and head of media and entertainment at Solomon Partners. “This has shortened the performance history needed to support securitization, allowing younger catalogs to be underwritten with greater confidence.”

He also cautioned that the accuracy of the data collection will depend on how the AI model is trained: “There may be a feedback loop between the recommendations it provides and the content being consumed, thereby concentrating consumption and exaggerating trend cycles.”

While it’s still the early days, the question of how AI collateral could potentially be treated in a bankruptcy or restructuring scenario is already being discussed by bankruptcy judges and lawyers, according to restructuring sources.

A restructuring veteran confirmed to 9fin that many lawyers are in agreement with the copyright law as outlined by FTI’s Gupta above, but whether any cases actually make it to the bankruptcy court remains to be seen.

“I think AI restructuring may be similar to that of crypto — you’re more likely to see a wave of companies acquiring peers and assets or liquidation scenarios, rather than AI companies filing for Chapter 11 bankruptcy,” the person said.

Stay tuned for Part Two of this AI feature, as we unpack how leveraged credits are implementing the very technologies that could ultimately undercut their businesses.