Music and entertainment transactional counsel.

The AI track moved on three continents at once. The US Copyright Office published its report in parts — digital replicas on 31 July 2024, copyrightability on 29 January 2025, and a pre-publication Part 3 on generative AI training on 9 May 2025. The EU AI Act entered into force on 1 August 2024 and its general-purpose AI obligations became applicable on 2 August 2025. The United Kingdom took longest: the Department for Science, Innovation and Technology opened its copyright and AI consultation on 17 December 2024, closed it on 25 February 2025, and published its report and impact assessment in March 2026.

Read those three dates as a drafting calendar rather than as policy. For fifteen months a London publisher could not finalize a standard-form AI training license, because the default it would be drafting against was unresolved. When the report landed, the work did not arrive gradually.

The commercial answer arrived on the same clock. The National Music Publishers’ Association announced industry-wide AI licensing pacts with Udio, finalized, and an agreement in principle with KLAY at its annual meeting on 11 June 2026; Suno announced a global licensing alliance with BMG on 12 August 2026 covering recorded and publishing repertoire, alongside watermarking and fingerprinting arrangements announced on 6 August 2026. Neither disclosed dollar terms. Both are the same instrument: a settlement and a forward license negotiated together, under litigation pressure, by a transactional lawyer.

That pressure is documented and dated too. Sony Music Publishing and Warner Chappell Music filed suit against Anthropic on 28 August 2026 in the US District Court for the Northern District of California, seeking up to $150,000 per work for willful infringement and up to $25,000 per instance of alleged removal of copyright management information across tens of thousands of compositions, on Music Business Worldwide’s August 2026 report; the American Federation of Musicians had sued Universal Music Group and Warner Music Group on 5 June 2026 in the Southern District of New York over recordings licensed to AI platforms without compensation or credit. The same company is therefore defending, licensing and settling on one fact pattern, and one in-house desk sits across all three postures.

California added the second track, and it is contract drafting rather than policy. AB 2602 took effect on 1 January 2025 and makes a contractual authorization for digital-replica use unenforceable unless the clause describes the intended uses with reasonable specificity and the performer either had counsel who separately negotiated it or was covered by a collective bargaining agreement addressing replicas. AB 1836, effective the same day, attaches statutory liability of a minimum $10,000 or actual damages to an unauthorized replica of a deceased performer. A Los Angeles company signing performer paper after that date needs a drafted clause on every agreement, not a template.

The federal answer is close enough to draft toward: the Senate Judiciary Committee advanced the NO FAKES Act unanimously on 18 June 2026, on the RIAA’s account of the vote. And on 10 July 2026 a coalition including IFPI, the RIAA, the Recording Academy, SAG-AFTRA and the Human Artistry Campaign launched a track-level AI labeling program, citing AI-generated tracks at 44 percent of new Deezer uploads and over a third on Apple Music. Labeling is metadata policy at the announcement and delivery-agreement language by the time it reaches a desk.

Read that line once and the structural problem is obvious. The floor of the disclosed law-firm band for a fifth-to-eighth-year music associate in Century City sits above the ceiling of the disclosed in-house band for a ten-year senior counsel at a streaming platform, on two postings read in the same city on the same day. This is an observed spread between two specific employers rather than a market average — two postings are not a survey. It is also the exact comparison a candidate makes.

London has no equivalent evidence, and no honest way to manufacture it. The only published London figures near this seat are firm-wide newly qualified rates at firms that happen to carry media practices: Mishcon de Reya moved to £110,000 effective September 2026 and Baker McKenzie to £150,000 in July 2026, both reported by Legal Cheek that year. Neither isolates an entertainment seat from the firm-wide rate. A London offer is therefore built from the general in-house market for the company’s size — our in-house counsel pay benchmarks carry that reference — and then moved by the reporting line, because a counsel reporting into a chief executive is priced as an officer and one reporting into a commercial director is priced as a manager.

The measurable part of the search is ours. Sartori has worked the London market for more than ten years and closed 24 in-house searches there across the trailing three years, with a 93 percent completion rate and a typical timeline of four to seven months. Our London mandate telemetry puts counter-offer incidence on that book at 32 percent and the median offer-to-acceptance window at 13 working days. Coverage is a separate measure: Sartori maps roughly 30,000 lawyers in London, inside a global program of about 1.5 million mapped profiles, thousands of mandate records and quarterly market surveys running since 2019.

Now the finding that does not flatter us. Of the six music, media and rights mandates inside those twenty-four, the 2 written as a single requisition covering catalog M&A, day-to-day licensing and AI training rights both ran past the seven-month end of our own stated band, and one closed only after the client agreed to split the work between a hire and a standing outside panel. We did not argue either client out of the combined brief at the outset, and we should have. A requisition that asks one person to be three specialisms is not a difficult search; it is two searches sharing a headcount, and the market answers it with candidates who are credible on one file and coached on the others.

A second limit is worth stating plainly, because it bounds half of this article. Our London telemetry does not reach Los Angeles. We hold no mandate record and no interview cohort for Los Angeles music and entertainment work, so every Los Angeles number here comes from employer postings and California statutes, read on the dates given. Our data sees one of the two cities this seat lives in.

Two engagements, anonymized and typical of the book.

A London catalog fund making its first legal hire. The brief was a general counsel to own acquisitions, administration oversight and the AI schedule, reporting to a chief executive. Nine weeks of market work produced a shortlist strong on corporate M&A and empty on rights administration, and we told the board the specification was the problem. Rewritten with administration instructed out and the AI schedule kept in-house, it closed inside the four-to-seven-month band with a lawyer moving from a rights holder’s own legal department. The complication was internal: the finance function had run the acquisition paper for two deals and read the new seat as a verdict on its work.

A European rights company with a streaming licensing book. The requirement was a licensing lawyer who could also hold an AI training-rights negotiation already under way. Candidates who read best on paper came from technology licensing and had never negotiated with a collecting society; the one who closed had, and took a flat title to get the AI file. A counter-offer arrived on day eight of the offer window and failed the way those usually fail.

The pattern under both is the sequence. The argument for this seat is strongest before the transaction and unarguable after the demand letter, which is backwards from the order in which a board releases the headcount — and the instruments in section 03 were public and dated long before either letter arrived.