Contained in the UMG-backed patent portfolio concentrating on AI music derivatives: A technical blueprint for the walled backyard mannequin?
The music trade is within the grip of an intensifying debate over how AI-generated derivatives of current music must be created, licensed, and monetized.
Common Music Group has been on the heart of that debate. The corporate has championed a “walled backyard” mannequin for AI derivatives, central to its settlement with Udio final October, the place AI-generated music can’t be downloaded or distributed outdoors the platform on which it was created.
Michael Nash, UMG’s EVP and Chief Digital Officer, has argued that with out such restrictions, AI derivatives danger permitting customers to “successfully use artists’ content material and their model to create derivatives the place you’re going to compete with the artist on different platforms.”
That place has put UMG at odds with AI music platform Suno, whose Chief Music Officer Paul Sinclair has advocated for “open studios, not walled gardens.”
UMG and Sony Music proceed to pursue litigation in opposition to Suno, whereas Warner Music Group struck a take care of the platform in November that preserved customers’ capability to obtain their creations.
In the meantime, the demand for AI by-product instruments is constructing from the platform facet. In February 2026, Spotify Co-CEO Gustav Söderström mentioned the streaming big’s expertise to let followers make AI-generated remixes and covers was “prepared,” however that “the absence of a rights framework” was holding issues up.
Towards that backdrop, an entity linked to UMG has been constructing a patent portfolio round AI-music infrastructure, via a partnership with IP asset administration, funding, and advisory agency Liquidax Capital. The expertise, relying on how it’s commercially deployed, might probably help a so-called ‘walled backyard’ strategy to AI-generated music derivatives, amongst different attainable purposes, which at present stay unclear.
In July 2025, UMG introduced a strategic partnership with Liquidax, led by CEO Daniel Drolet, to speed up the event and licensing of music-related AI patents.
A brand new entity, Music IP Holdings, Inc. (MIH), was fashioned to license the expertise into the worldwide market. UMG mentioned on the time that it had filed 15 patents with Liquidax throughout fields together with musical collaboration, multimedia content material creation, AI risk safety, and rights administration.
By November 2025, MIH had opened a headquarters on Nashville’s Music Row, with Drolet serving as Chairman and CEO. In a press launch asserting the growth, the corporate mentioned it held “greater than 60 protected improvements with quite a few further expertise households and portfolios beneath improvement” — a major growth on the 15 filings disclosed 4 months earlier.
MBW has obtained three of these patent filings and reviewed them intimately. The filings are assigned to MIH, to not UMG straight, and neither MIH nor UMG has publicly disclosed whether or not or after they intend to make use of or license the described expertise, or to whom.
Right here’s what the filings describe…
‘AI-GENERATED MUSIC DERIVATIVE WORKS’
This submitting seeks to deal with the core copyright drawback posed by AI skilled on and used to remix current music.
The invention the patent claims is a system constructed round artist approval. A person asks the AI to rework a copyrighted observe, the system checks whether or not the rightsholder would approve, and if it passes, the output will get a digital watermark and is launched beneath phrases the rightsholder has set.
Filed on October 24, 2024 and granted on June 3, 2025, that is the earliest-granted patent within the household. It initially listed Daniel Drolet as the only inventor.
A subsequent expanded continuation, assigned to MIH and at present pending, was filed in August 2025 and provides Chris Horton, Jeremy Uzan, and Sion Elliott as co-inventors alongside Drolet. The three are recognized on LinkedIn as EVP, Strategic Expertise at UMG; Director, AI and Superior Expertise at UMG; and Director, World New Enterprise Technique at Common Manufacturing Music, respectively.
You may learn the continuation right here and the unique granted patent right here.
The summary within the up to date continuation submitting describes “a system and technique for creating AI-generated by-product works from predetermined content material with copyright compliance and content material proprietor management”.
It provides that “in some facets, the system receives predetermined content material and a user-requested transformation theme, then employs generative synthetic intelligence to create a by-product work. The system could allow scalable rights administration for AI-generated content material throughout music, video, textual content, and different media codecs.”
It describes by-product works as “creations which are primarily based on a number of predetermined works” and warns that AI-generated variations “could infringe on the rights of the creator or copyright holder of the predetermined work, via copy and transformation with out permission” and “could also be adversarial to preferences of the creator or copyright holder of the predetermined work.”
Within the submitting’s personal language, the tactic covers “receiving a request to rework the predetermined content material right into a by-product work, receiving a requested theme for the by-product work, utilizing generative synthetic intelligence to create the by-product work generated as a perform of the predetermined content material and the requested theme, figuring out if the generated by-product work is permitted primarily based on a machine studying mannequin configured to find out a content material approval rating as a perform of content material proprietor preferences, in response to figuring out the generated by-product work is permitted, making use of a digital watermark to the permitted by-product work, configuring an authorization server to control use of the permitted by-product work primarily based on the digital watermark and offering person entry to the licensed by-product work.”
The person’s directions to the AI might be gathered conversationally. The submitting states that “the requested theme could also be decided utilizing a Massive Language Mannequin (LLM) and a chatbot interview.”
Watermarks may also expire, giving rights holders a solution to handle how lengthy a by-product stays licensed. The submitting states that “the authorization server could also be configured to revoke, or not renew, approval for a by-product work, primarily based on permitting the watermark embedded in a by-product work to run out.”
‘MULTI-STAGE APPROVAL AND CONTROLLED DISTRIBUTION OF AI-GENERATED DERIVATIVE CONTENT’
Filed on Could 2, 2025 and granted on September 23, 2025, this submitting, which you’ll learn right here names Chris Horton, Jeremy Uzan, Sion Elliott, and Daniel Drolet as inventors. It builds on the earlier-granted patent, with broader claims.
The place the sooner patent establishes the bottom system, this one tackles a special drawback: learn how to construct approval logic that displays the artist’s personal preferences at each ends of the inventive course of.
The submitting describes a way that mixes a machine-learning strategy with “predefined rule units established by content material house owners that specify permissible and/or impermissible transformations for particular content material” — the 2 can function “both individually or together, to accommodate totally different content material proprietor preferences and use instances.”
The core addition is a two-stage approval course of: one verify earlier than the AI generates something, one other verify on the completed output.
The submitting refers to “pre-generation desire information” and “post-generation desire information” related to a “content material authority” — outlined broadly as “any entity having reputable management, decision-making energy, or governance rights over digital or inventive content material and its permissible transformations,” together with “a pure particular person (reminiscent of an artist, creator, or designated particular person)” in addition to authorized entities and rights administration organizations.
The submitting provides a concrete instance of how the approval system displays artists’ private values: “If an artist is vegetarian and doesn’t need their voice or model for use in songs about meat consumption, this desire might be captured by their label forward of time and offered to the content material derivation platform. This permits filtering to happen each on the immediate degree and on the output degree, making certain alignment with the content material proprietor’s values all through the creation course of.”
“If an artist is vegetarian and doesn’t need their voice or model for use in songs about meat consumption, this desire might be captured by their label forward of time and offered to the content material derivation platform.”
excerpt from Music IP Holdings patent submitting
Artist consent can be constructed into voice-substitution claims inside the submitting. The submitting describes a function that lets customers “substitute the lead vocals in an current recording with the voice of one other artist who has consented to such transformations” — and notes that “the choice motion itself could set off the attribution and remuneration system, which ensures applicable compensation to each the unique and substitute artists.”
Rejected requests can include an evidence. The submitting describes “a remediation report when a by-product work fails to fulfill approval standards”, analyzing what prompted the failure so the person can strive once more.
The patent additionally units out how distribution controls can stop unauthorized use of derivatives on third-party platforms. Associate platforms together with streaming companies, social networks and distribution aggregators “could also be contractually required to scan incoming content material for the presence of such markers, and take automated enforcement motion primarily based on the encoded guidelines.”
The submitting additionally describes “context-restricted playback, the place remodeled works are solely accessible inside the permitted media surroundings” — one in every of a number of distribution configurations the patent covers, however one which sounds loads just like the ‘walled backyard’ strategy championed by UMG. Not one of the filings use the time period ‘walled backyard’.
The submitting additional describes automated income distribution, with “a wise contract that routinely allocates income from every licensed distribution of the by-product work to stakeholders recognized within the utilization registry”, triggered “every time the digital identifier is verified throughout a streaming session.”
The rights framework the patent describes has components in widespread with the type of infrastructure Söderström recognized as at present lacking — although the submitting makes no reference to Spotify or every other platform, and MBW has no info on industrial discussions between MIH and any streaming service.
‘AI-GENERATED DERIVATIVE CONTENT SCALING FOR MERCHANDISE’
Filed on October 3, 2025, this pending submitting names Horton and Uzan as inventors.
This submitting, which you’ll learn right here, tackles a special industrial drawback: learn how to prolong the identical approval-and-watermark system past audio into bodily and digital merchandise. It applies to “AI-generated album art work derivatives, attire designs, posters, and digital items”, all beneath the identical governance framework because the audio filings.
The fan-facing facet of the system is designed round a five-step journey on a cellular app. The fan first picks a supply asset — lyrics, moments from a track, or themes linked to an artist. The AI then generates a set of design variations. The fan picks one, customizes it by adjusting shade, measurement, and place, and sees a real-time preview rendered both as a bodily product or in augmented actuality. On the ultimate step, the fan chooses to purchase or save.
Behind the app sits a backend that pulls from a library of artist reference supplies — artist logos, artist fonts, lyrics and textual content, album artwork, and tour visuals (see illustration under). That is the curated set of inventive belongings the rightsholder makes accessible for followers to remix.
The fan’s request passes via two AI parts. A “information agent” checks it in opposition to the rightsholder’s model pointers. An “motion agent” then runs the generative AI and produces the design. The design then goes via an approval verify: if it fails, it’s rejected (a T-shirt studying “Music is simply noise!” within the submitting’s personal illustration); if it passes (a T-shirt studying “Stage is my canvas”), it strikes on to achievement.
From there, the system splits in two. Bodily merchandise is routed to a “print agent” that feeds a print-on-demand accomplice, and a “store agent” that lists the product on an e-commerce storefront. Digital merchandise takes a separate path to a 3D rendering engine and metaverse platforms — the place, within the submitting’s illustration, an avatar is proven carrying the AI-generated design inside a digital surroundings.
Each tracks feed right into a contract that distributes income to stakeholders. One permitted design can grow to be a bodily T-shirt, an e-commerce itemizing, and a wearable merchandise inside a metaverse — with automated royalty funds routed on the finish.
The submitting additionally describes the platform working in actual time throughout dwell occasions, with concert-synchronized merchandise technology pulling in information concerning the track being performed, stage visuals, lighting, crowd engagement, and venue location to provide context-specific gadgets. Fan co-creation interfaces verify every modification in opposition to model guidelines in actual time. Authentication combines bodily components like holographic tags or QR codes with embedded digital identifiers.
The three publicly accessible filings reviewed right here symbolize solely a fraction of what MIH says it’s constructing.
The corporate’s declare of greater than 60 protected improvements suggests the portfolio extends properly past the by-product works household into the opposite fields UMG has referenced, including musical collaboration, music and well being, and AI risk safety.
UMG’s Michael Nash publicly referenced the corporate’s AI patent exercise in remarks on the HumanX convention on April 8.
“With respect to leaning into innovation too, to reveal to artists that we’re centered on directing our sources to additionally create options, we’ve developed numerous AI patent purposes,” Nash mentioned. “One good instance could be when it comes to well being and wellness. We developed a course of that’s enabled by AI which helps our unique participation within the Sound Remedy class on Apple Music.”
MBW has beforehand reported on a separate UMG patent submitting protecting AI-generated binaural beats.
On UMG’s Q2 2025 earnings name final July, Chairman and CEO Sir Lucian Grainge famous that the expertise underpinning Sound Remedy is just not the one tech constructed in-house utilizing AI.
He added that the corporate had been creating AI-powered expertise since 2020 “to help artist advertising, analytics and distribution”.
He described the Liquidax partnership as the following step: “To speed up and scale the event of our patents, we not too long ago partnered with Liquidax Capital, an IP asset administration investing advisory agency,” Grainge mentioned, acknowledging that “Liquidax, on our behalf, has already filed 15 patents within the fields of musical collaboration, multimedia content material and marketing campaign creation, AI risk safety, music administration and rights administration.”
He added that UMG’s “enormously expanded patent portfolio can then grow to be a catalyst to speed up introduction of merchandise to {the marketplace}.”
UMG has declined to remark additional, and the complete plan for the tech described within the filings above stays unclear.Music Enterprise Worldwide
