Introducing Studio Day

Theme

At the end of August, the music publishing arms of Sony and Warner sued Anthropic, alleging it used hundreds of copyrighted song lyrics and pieces of sheet music to train Claude. The case is part of a growing fight over what AI companies can train on. It also points to a broader shift. Creative archives have gained a second use. Alongside storing work that can be seen, heard, or licensed, they can now train systems to make new work. For brands, old campaigns, images and design files can teach a model what the brand looks and sounds like, then help generate what comes next. Ownership is one part of the question. The next is who decides what the machine should learn from the archive.

Signal

The data that goes into a training set matters, because a model will repeat the patterns it sees most clearly. If an archive overrepresents one campaign, era, or visual style, that version of the brand becomes easiest to reproduce. For instance, when the fashion brand Coach used Adobe Firefly Custom Models, their team curated images of existing handbags into separate training sets for specific products and styles, teaching the system recognizable "Coach Codes" such as its hang tag, materials, and hardware. That curation helped the models produce accurate digital twins while giving designers room to explore new variations.

We recently faced the same question in a different context while extending the Artist and the Machine brand guidelines to become both human and machine-readable. Fonts, colors and repeatable rules were easy to define, while taste, context, and knowing when to break the rules remained more challenging to encode. For brand leaders, the work begins before anything is generated and remains uniquely human. Once that foundation is in place, AI can help teams iterate and expand far more efficiently while preserving what makes the brand distinct.

Introducing Studio Day

Summit is where ideas spark. Studio Day is where we make things real together.

New this year: Studio Day LA 2026 — Thursday, December 3, Audrey Irmas Pavilion, the day after Summit. Invitation-only: six hand-curated Studio sessions (creative challenges, salons, masterclasses), Jeffersonian private lunches, and a closing Creative Showcase. Leave with something you learned and something you made. The room: leaders from Apple, Netflix, Disney, LVMH, NVIDIA, OpenAI, and beyond, with CapCut, Adobe for Startups, Luma, and Artlist as partners.

Missed early bird tickets for Summit? Get a standard tickets while they last: Apply for Summit

Pulse

• Scanned, Shredded, Subpoenaed: Music publishers allege Anthropic bought songbooks, scanned every page, and discarded them — and now want the court to reveal exactly which songs went in. The archive fight is moving from did you train to on precisely what.

• Consent, Shipped as Product: Believe and TuneCore will require artists to opt in twice — to AI programs generally, then to Suno by name — before a note reaches a training set. "The artist must consent first, period."

• Brand Guidelines Learn to Read Themselves: Markup AI expanded its Content Guardian Agents into Word, ChatGPT, and Claude — software that checks generated content against brand guidelines in real time. The machine-readable brand book is becoming a product category.

If you’re creating something aligned with Artist and the Machine or you’d like to recommend work shaping this space, we’d love to see it for the chance to be featured. Reach out to us at [email protected].

Til next time,

Artist and the Machine.

xx