If your team uses AI to draft marketing copy, product content or creative variations, the hard part isn't getting words on the page anymore. It's knowing when a helpful draft crosses into something you shouldn't publish.
That gets riskier when employees ask a model to mimic a known artist, rewrite a famous song or produce something "close to" existing work. The output may look like a productivity win right up until legal, brand or partner concerns show up.
Elvis rights holders widen the copyright fight
Authentic Brands Group and other Presley rights holders sued Anthropic, alleging Claude was trained on copyrighted Elvis lyrics without licensing and still produces infringing outputs or unauthorized adaptations. Anthropic told NBC News it disputes the claims and will defend the case.
The complaint seeks damages tied to 274 songs and asks the court to stop further infringement. Whatever the case outcome, it shows how quickly AI use can move from experimentation to intellectual property exposure.
- Training claim. Elvis Presley rights holders allege Anthropic used lyrics from 274 songs without permission to train Claude models.
- Output claim. The lawsuit also says Claude can generate infringing lyrics and close adaptations of Presley songs when prompted.
- Money and remedies. Plaintiffs are seeking $41.1 million in statutory damages, plus an injunction and destruction of unauthorized copies.
- Part of a bigger pattern. NBC News places the case in a broader run of music industry lawsuits against Anthropic over lyrics and song-related infringement.
Why this matters beyond music
Most businesses aren't training foundation models on song catalogs. But many are using AI to create public-facing material, and that creates a simpler, more common problem: employees can ask for work that is too close to someone else's protected content.
The real operational issue is policy drift. One person may know not to request a rewrite "in the style of" a famous creator, while another treats that as a normal shortcut under deadline pressure. Without clear guardrails, risky prompts and risky outputs spread quietly.
This also isn't just a legal team issue. Brand, sales, marketing and agency teams all use AI in fast-turn workflows where review can be light. If the rules live in a memo nobody reads, they won't shape the work when it matters.
The safer approach is to move from vague warnings to system-level boundaries. Decide what kinds of requests are off-limits, where human approval is required and which source materials your team should rely on instead.
How Metomorph can put boundaries around creative AI use
Metomorph is useful here because it lets a business turn policy into day-to-day workflow. You can set business rules in plain language, attach approved material in a knowledge base and keep the work inside a defined project instead of scattered across personal chats.
Imagine your marketing team preparing a seasonal campaign with AI help. In that project, you could load brand guidelines, approved past campaigns and legal notes into the knowledge base so drafts are grounded in your own material rather than whatever the model happens to recall.
Then add business rules such as: do not imitate living artists or named songwriters, do not reproduce copyrighted lyrics, and route externally published creative copy for approval. Those rules can constrain work across chats, skills, forms and apps when configured.
A team lead could package the approved campaign drafting process as an organization skill, so staff start from the same method each time instead of improvising prompts. That doesn't guarantee perfect output, but it does reduce variance and makes review more consistent.
For higher-stakes requests, multi-model consensus chat can give your team a more defensible first draft by comparing several frontier models at once. Used with your own knowledge base and rules, that helps shift AI from open-ended mimicry toward controlled business content.
Practical steps this week
- Write three plain-language AI content rules now. Start with what your team must not ask for, what it must not publish and what needs approval before release.
- Ground creative drafting in your own documents. Load brand guidance, approved copy and policy notes where your team actually works so people stop improvising from memory.
- Create one standard workflow for public-facing AI content. If different employees use different prompt habits, your risk is already uneven.
- Pick one active content project and add explicit approval steps for anything that references third-party creative works or styles. That's a concrete place to start this week.
Explore Business rules, Knowledge base, Organization skills in Metomorph.
Source: NBC News, Elvis Presley’s rights holders sue Anthropic, saying it trained AI on his music (October 25, 2026).