Recording a document of transfer with the US Copyright Office goes from $95 to $215, which buys the recordation of exactly one work. Everything past the first work sits on a five tier escalator nobody has reported, which means an assignment is now priced by the length of its works schedule. Elsewhere this fortnight, the Code of Practice that settles what machine readable marking of AI generated audio actually consists of has 95 provider signatories without a single record label, publisher, distributor, streaming service or society among them. PRS will begin taking exact play data from around 7,000 UK premises next month into a matching process whose published inputs are a title and a writer name. The theft of Eminem's unreleased catalogue was caught because somebody recognised a file list. Four stories, one question underneath all of them, which is who keeps the list of what you own.
Policy & Regulation
The Fee For Recording A Catalogue Transfer Is Now Set By How Many Works Are In It
The Copyright Office's proposed fee schedule has been sitting on a public server since the summer while the music trade has taken one number out of it. Electronic recordation of a document of transfer rises from $95 to $215, which is the figure in every carrier of the story. Open the schedule at the Section V.A chart and that $215 covers the recordation of a single work, identified by one title or one registration number. Everything beyond that one work is charged separately, which is where the cost of moving a catalogue now lives.
The add on for additional works and alternate identifiers runs on five tiers. One to fifty goes from $60 to $215. Fifty one to five hundred goes from $225 to $300. Five hundred and one to a thousand goes from $390 to $520. One thousand and one to ten thousand goes from $555 to $745. Above ten thousand the proposed structure is $745 per additional group of ten thousand works. Each further transfer recorded inside the same document is another $215, up from $95. Paper recordation goes from $125 to $350, with a paper notice of termination on Form TCS at $275. Put those together and recording an assignment of a two thousand work catalogue is $215 for the document plus $745 for the works, against $95 plus $555 today.
Recordation under 17 U.S.C. 205 is voluntary, which is exactly why unrecorded assignments accumulate in independent catalogues for years at a time. What the new schedule does is attach a price to the thing most labels treat as an appendix. A works schedule used to be whatever came out of a spreadsheet on the day of the deal, pasted into an exhibit and never read again. On these tiers it is a priced object, so the cheapest version of any recordation is the one where the title list is clean enough to file once rather than twice.
The schedule used to be the attachment. On these tiers it is the invoice.
The second half of this is the public record you are paying into. The Office updated its bulk registration and recordations dataset in January, the first substantial update since 2022, covering registrations from 1 January 1978 to 27 June 2025 and no further. A recordations file is published as raw parsed data. The only tabular recordation output is a terminations CSV, with the rest of the recordation tables marked "more coming soon" and registration cancellations likewise. The read me documents the data as MARC 21 served as record_id, tag, subfield, value and row, documenting fields for author and claimant information while documenting nothing for the parties to a transfer or for a recordation document number. The Office's own warning on the datasets page is that the set "should not be relied on for legal matters".
So the position for a publisher checking whether an assignment in its own chain of title was ever recorded is this. The fee for doing it properly is about to be set by the size of the asset. The bulk record of what has already been recorded stops roughly sixteen months before the change, publishes no documented field for the assignor or the assignee, then carries a disclaimer against being relied on. The answer has to come from your own files, which means the practical consequence of this fee schedule is not a budget line. It is that the only authoritative register of your own transfers is the one you maintain, where having it right has just become cheaper than having it wrong.
One date discipline, because it matters this week. The schedule itself states no effective date. The cover letters are dated 14 July 2026 and say the Office "seeks to implement these new fees in the fall of 2026" and that the Register may institute them "120 days after the proposed schedule is submitted to Congress". One hundred and twenty days from 14 July is 11 November, which is arithmetic rather than a published date, so the secondary carriers reporting 11 and 12 November are reading the clock rather than quoting the Office. This issue publishes on 10 November. Anything recorded electronically before that window closes clears at the current figure.
Sources: US Copyright Office, "Proposed Schedule and Analysis of Copyright Fees to Go Into Effect", read at source for the Section V.A fee chart and both cover letters · Federal Register, "Copyright Office Fees", 20 March 2026, the underlying proposed rule · United States Copyright Office Datasets, read at source for the coverage dates, the published files and the disclaimer · The dataset read me, for the MARC 21 structure and the documented fields · Copyright Office NewsNet Issue No. 1080, 22 January 2026, for the dataset update · PerspireIP, 14 October 2026, which is the carrier working the statutory clock
Technology
Nobody From Music Turned Up To Decide What An AI Marked Audio File Looks Like
The European Commission's Code of Practice on Transparency of AI generated Content is the one place where the question of what counts as machine readable marking is being worked out in operational detail rather than in principle. Its working group text commits signatories to outputs "marked in a machine-readable format and detectable as artificially generated or manipulated" using solutions that are "effective, interoperable, robust, and reliable as far as technically feasible". Audio is expressly in scope alongside image, video and text. The Commission's own page carries 95 signatories in Section 1, being providers, against 192 in Section 2, being deployers.
We read both lists name by name. There is no record label on either of them. No music publisher, no distributor, no streaming service, no collecting society, no AI music generator. Google, Meta, Microsoft, OpenAI, Anthropic, Mistral and Cohere are Section 1 providers. So are Synthesia, Resemble.ai and Getty Images, which appears in both sections. Getty turned up for the image trade. The only music suggestive names anywhere in either list are Hastings Music GmbH and a company called Label4ai, neither of which the Commission describes, so we are not going to pretend to know what they do.
The convention for how a generated recording declares itself is being written by model providers and detection vendors. The first a label will know of it is when the files start arriving.
Adherence to the Code is voluntary. The Article 50 transparency obligations it is built to help demonstrate compliance with are not, the Commission's page stating that those obligations have applied since 2 August 2026. The Commission and the AI Board have confirmed the Code as an adequate voluntary tool for showing compliance with them. Which means the sector that will eventually be asked to preserve, deliver and audit a marking field has sat out the exercise that defines the field, in a period when the obligation was already live.
It fits an uncomfortable pattern. Obligations attaching to provenance metadata are arriving through documents a label accepts by carrying on using a product rather than through standards it is consulted about. One law firm's clause level reading of Meta's pending commercial terms, which take effect on 20 November, records an AI Features section barring a business customer from altering or removing "any watermark or content-authenticating metadata". We have not read Meta's own preview pages, which refuse automated access, so that clause is a law firm's rendering rather than ours. Taken with the Code, the direction is the same from both ends. Somebody is deciding what a marked file has to carry. Neither decision is happening anywhere a label would think to look.
Sources: European Commission, "Strong backing for the Code of Practice on Transparency of AI-generated Content", published 31 July 2026 and last updated 24 September 2026, with both signatory lists read at source on 8 November 2026 · European Commission, Code of Practice policy page, for the working group commitments, the voluntary status and the 2 August 2026 date · Shumaker, Loop & Kendrick client alert, 21 October 2026, for the clause level reading of Meta's pending commercial terms
Metadata & Standards
Seven Thousand UK Premises Will Report Exactly What They Played Into A Process That Matches On Title And Writer
PRS for Music has agreed a partnership with Soundtrack, the business music service, under which it will receive data showing the exact music played across around 7,000 UK premises. Soundtrack puts its repertoire at 150 million songs. PRS begins incorporating the data from December 2026. Every carrier has written this as an accuracy story, which it is. The interesting half is what happens to that data after it arrives.
Start with what PRS actually said, because the first version of this story in the trade press overstated it. The society's line is that the partnership "reduces reliance on analogous data where actual usage information has historically not been available". That is an addition to the existing method rather than a replacement of it. From December a subset of UK public performance usage will arrive as exact match data while the rest will not, so the same work will be identified by two different mechanisms depending on where it was played.
Then the matching. Neon Music's account cites PRS's own published policy as matching reported usage against its database "using details such as the ISWC, PRS tunecode, title and writer names". Neither PRS nor Soundtrack has published anything about what the usage file itself carries. Record of the Day's report of the announcement does not mention ISWC, ISRC or any identifier at all, nor does anything else we could read. That is our observation rather than either party's statement, yet it is the whole operational point. Exact play data is only as good as the key it joins on. A work whose ISWC is absent, or whose writer strings do not match the ones in the register, is in precisely the position it was in under the sample.
Better usage data does not fix a matching problem. It makes the matching problem visible at a higher resolution.
For an independent publisher the job this creates is dull and finite. Establish which of your works carry an ISWC in the society's register rather than in your own spreadsheet, then establish which writer name strings you have filed against them. Exactly the same holds on the recording side, where ISRC validation is the difference between a usage report that lands on your repertoire and one that lands in a suspense file. Neon Music records two further things PRS has not published, being whether Soundtrack plays appear as a separate statement line plus which performance period the first December run covers. The third thing nobody has published is the one that decides whether any of it reaches the right work.
Sources: Record of the Day, 28 October 2026, read at source for the 7,000 premises, the December start, the 150 million song repertoire and the PRS quotation · Neon Music, Alex Harris, 2 November 2026, read at source for the published matching inputs and the unpublished points · Digital Music News, 27 October 2026, for the named premises on the Soundtrack platform
The Cautionary Tale
The Theft Of Eminem's Unreleased Catalogue Was Caught Because Somebody Recognised The File List
On 22 October the Department of Justice announced that Joseph Strange, 48, of Holly, Michigan, had pleaded guilty to criminal infringement of a copyright in the Eastern District of Michigan. He had worked for Marshall Mathers from roughly 2007 until 2021 and copied, in the department's words, "10 or more audio files" from a studio in Ferndale, Michigan. He sold them between July and August 2024 for approximately $50,000. The files had been accessed more than eight million times by November 2025. The government puts the infringement amount at at least $750,000. Sentencing is set for 27 January 2027.
The mechanism is in the 2025 charging release rather than in the plea. On or about 16 January 2025 the FBI was contacted by employees of the Ferndale studio, who had "recently discovered unreleased music created by Mathers that was available on the internet". What they had found, in the release's own sequence, was that they "obtained an image of a list of the music that Mathers had created, but not released" and that was "for sale through various internet sites", then "recognized this as an image taken directly from a hard drive in the Ferndale studio". The theft was identified from a register of what the catalogue contained, not from the audio. Purchasers then named the seller.
Nothing in either release describes a watermark, an access log, a metadata check or an audio fingerprint doing any of the work. The person who had helped build the storage had standing access to it for fourteen years, which is the ordinary condition of most independent archives rather than an exceptional one. The only thing that made the material identifiable as having left the building was a list of what was in the building. A label that cannot produce that list on request cannot establish what it lost, which is a different and much worse problem than losing it.
Sources: US Department of Justice, Office of Public Affairs, 22 October 2026, read in full at source · US Attorney's Office, Eastern District of Michigan, 19 March 2025, read in full at source for the discovery sequence and the counts originally charged. Neither release cites a statute section, names a judge or gives a case number. The March 2025 complaint also charged interstate transportation of stolen goods, which the October release does not mention, so nothing here should be read as resolving that count.
On The Wire
Four shorter things. None of them reported in the music trade press.
Metadata & StandardsMusicBrainz has moved its login and set the date every existing OAuth application stops working
If you run catalogue reconciliation against MusicBrainz, which in practice means anyone cross checking an identifier or a work title against the largest open music database, there is an authentication migration in your diary whether you have put it there or not. MetaBrainz posted in June that accounts and authentication were moving, with the migration itself shifted from 6 July to 27 July 2026. Base URLs change from musicbrainz.org/oauth2 to metabrainz.org/oauth2, with the instruction to "update these base URLs in your code". The applications page moves. Existing applications "are deprecated" and "will not be migrated to MetaBrainz OAuth". Proof Key for Code Exchange is now mandatory for all flows. The date on the back end of it is 1 August 2027, when digest authentication and MusicBrainz OAuth applications "will be removed", at which point existing applications "will stop working". That is the second dated break in this database inside a year, the other being the search service upgrade of 30 November, with neither covered anywhere in the trade press.
Read Across748 licensed tracks are expiring one at a time and the only published record of which one goes next is a fan site
Rock Band 4 came off the PlayStation and Xbox stores on 5 October 2025 on the game's tenth anniversary, Harmonix's community manager Kyle Wynn explaining that "the original licenses for the core soundtrack are expiring". Video Games Chronicle put the catalogue at 748 tracks. The downloadable content did not go with it. It has been coming off one track at a time ever since, on what one fan site reckons are ten year licence terms, hedged in its own text as "seemingly", so the sequence of expiries is an inference from the dates rather than a published schedule. Which it is, because nobody publishes one. The record of what leaves next is a weekly list on gamersocialclub.ca, where the entry for 3 to 9 November 2026 is a single track, "Born Again Tomorrow" by Bon Jovi, scheduled to go on the 3rd "or thereabouts", with the caveat that "some DLC will get pulled without warning, so if there's a track you really want, go get it!". The read across is exact and not funny at all. A licence term is a date held in somebody's contract file, so where nobody maintains a schedule of those dates the only way to discover an expiry is to watch the shelf. The rights holders who licensed into that catalogue are in the same position as the buyers.
Streaming PlatformsOne work, 6,231 views, with a platform's commercial library doing the job of a licence record
Matthew Michael Hobbs filed a copyright complaint in the Central District of California on 19 October over "Cheese Tax", first published 28 February 2023 and registered on 13 March 2023 as SR 973-739. The pleaded facts are small. One Kroger TikTok post dated 10 March 2023, ten days after first publication, with at least 6,231 views, taken down on or about 23 April 2026 after the writer found it in December 2025 and gave notice in March. The mechanism is the part no carrier has reported. The complaint pleads that the works "are not, and have never been, available in TikTok's Commercial Music Library" and that the audio was taken from the writer's own user uploaded original sound at the handle @puppysongs. Kroger's position, as the complaint describes it, was that the works were available to TikTok business account holders, which the writer answered by supplying documentation that the audio was a user upload rather than a commercial sound. Membership of a platform's commercial library is therefore functioning as the entire licence record for brand use, with no field, flag or lookup telling a commercial account which of the two kinds of audio it has reached for. Any recording that sits on the platform without sitting in the library is exposed to exactly this use and will be defended on exactly this ground.
Label OperationsA UK music company went into liquidation in September. In October its director gave notice he intends to keep trading under its name.
HARP MUSIC AND SOUND LTD, company number 13333854, entered creditors' voluntary liquidation on 15 September 2026, with Gazette notices for the winding up resolution and the appointment of liquidators both published on 22 September. On 12 October a further notice appeared, for re use of a prohibited name. The director, Daniel Lidchi, gave notice of a new vehicle, Lidchi Creative Ltd, intending to trade as Harp Creative, Harp, Harp London, Harp Music and Sound, plus Harp Music & Sound. Five trading names in one notice. That route exists precisely because the name is the thing creditors recognise, which is the same reason it is the thing a counterparty record stores. Anybody holding a contract, a licence or a catalogue counterparty entry that says "Harp Music and Sound" now has two entities it could mean, one of them in liquidation and one of them a live business, with the only document connecting them a notice almost nobody reads. A counterparty field holding a trading name instead of a company number cannot tell the difference. This is the argument for the company number, made from the other direction.
And Finally
A Cheerleading Company Sought A Restraining Order Over White Boots. One Email Ended It.
Pro Silver Star, which owns the Dallas Cowboys Cheerleaders, petitioned on 26 October for a temporary restraining order against My Chemical Romance and Live Nation Entertainment over a performance at the Hollywood Bowl five days earlier. The pleaded claims were trade dress infringement, trademark dilution, unfair competition plus contributory and vicarious infringement. The asserted trade dress was uniforms with white boots, white shorts and a blue and white top with blue stars. Judge Christine Nowak then cancelled the hearing, giving the reason as "agreement reached". The agreement was an email from the band's lawyers, quoted in the claimant's own declaration, which read: "I write to let you know that there's no need to proceed with the petition for a TRO", continuing that "The Black Parade shows are finished and My Chemical Romance will not be performing any remaining shows on its current tour".
This is a trade dress claim rather than a music rights claim, which is the first reason it is in this slot and not further up the page. The second is the sentence that ends the report, which is that "It was not immediately clear what the development meant for the action pending against Live Nation". The costumes stopped. The matter did not.
There is a dull point underneath the boots and it is on theme. A touring production is full of rights bearing assets that nobody schedules: wardrobe, choreography, staging, projections, the lot. None of them appear on a delivery, so none of them appear on a register, so the first time anybody establishes what a show actually contains is when a claimant describes it in a pleading. The asymmetry is the joke. A registered trade dress in a cheerleading uniform was enforceable against a stage show inside five days. Ask the same production to evidence the clearances on the music it played that night and you would be waiting rather longer than that.
Source: Michelle Casady, The Texas Lawbook, "Facing Dallas Cowboys Cheerleaders' IP lawsuit, My Chemical Romance quickly concedes", syndicated 30 October 2026, read in full at source. The report gives no case number and does not clearly identify the court, so neither is stated here.
Feature Spotlight
DataDoktor: The List Is Only Worth What It Costs To Prove
Read the issue back and the decisive document in every story is a list rather than a work. A fee schedule that prices a transfer by the number of works on its schedule. A bulk public record with no documented field for the parties to a transfer. A matching process whose published inputs are an identifier, a title and a writer name. A commercial music library standing in for a licence record. A stolen catalogue identified by a register of what was in it. A delisting sequence that exists only because a fan maintains a spreadsheet. In each case the right was real. What decided the outcome was whether anybody could produce the list.
DataDoktor is the part of rightsHUB that reads the list back to you before somebody else does. It scans a catalogue against distributor, platform and society requirements, validates ISRC, ISWC, IPI, CAE plus ISNI, then flags values that are missing, malformed or duplicated. It checks contributor credits, artwork and audio specifications, genre tags, explicit flags, then DDEX and CWR compliance. What comes back is a prioritised action list with the correct value against each item, bulk fixes you review before they apply, a re scan, plus a catalogue quality score that moves when you fix something. Duplicate identifiers on one recording are the most common failure an independent catalogue carries, yet they are invisible until the moment a usage report or a works schedule has to join on them.
The question worth asking is the one this issue answers badly six times over. If a registry, a society, a buyer or a court asked you today for the list of what you own, with an identifier against every line, how long would that take and how many lines would come back wrong. The rest of the toolkit is at rightshub.net/features. A demo takes half an hour.
A catalogue is not what you own. It is what you can list, identify and hand over on the day somebody asks.