John Shiloh Beaumont II introduces Jack Calder and Diane Collins — hypothetical, composite characters built from real cases, real patent filings, real DMCA records, and the lived experience of practicing attorneys. Their names are borrowed, with no apologies to the cougar. Their problems are not fictional.

Jack's copyright problem. A friend sends him "Have I been trained?" — and he wakes up to 47 of his songs sitting in LAION-5B, a dataset of 5.85 billion image-text pairs that also links to millions of audio tracks. LAION didn't license anything. They scraped the open web. The DMCA removes the data from the public dataset, but it cannot untrain the model — the meal has already been eaten, the pattern is already in the bones, and the bones belong to a corporation.

Jack's patent problem. His fourth-generation guitar pickup is rejected for prior art — filed eight months earlier by Aurelia, an AI shell company that pre-invents 5,000+ applications a year, locks up future ideas, and demands tolls. A challenge costs $400,000 minimum. Examiners average less than 20 hours per application. And AI-made discoveries are automatically in the public domain — so the more we lean on the machine, the less there is left for anyone to own. "IP has cancer. The diagnosis is terminal."

Diane's war. A solo practitioner of 18 years, she subscribed to an AI document review service that flagged a two-year non-compete as boilerplate. "94% accurate" means six mistakes per hundred contracts — six expensive, career-ending mistakes. The big firms grab the headlines. The real damage lands on the solo practitioners who can't negotiate with the AI vendors.

The fix is not in the courtroom. It's in legislation — which means you need to talk to your legislator. The machine has no judge. But you do.

The legislator contact tool is live at beaumontandsheridan.com/act — enter your zip code, get your senator and representative and pre-written letters about AI regulation, training data, transparency, opt-out rights, content labeling, and data center accountability. One click opens your email client pre-filled. This is not a petition — it goes direct to their inbox. And inbox math is real: fifty emails and their legislative director notices. Five hundred and they schedule a meeting. Five thousand and they change the vote.

Learn more:

• Resource Library — guides on AI training data, having your work in a dataset, DMCA takedowns, and documenting infringement
• Is Your Work in a Dataset? — how to check whether your creative work was fed into the machines
• Contact Your Legislators — the live tool from this episode

Featuring Heavy Comforter music throughout. HeavyComforter.com — the real thing.

John Shiloh Beaumont II is joined by technology consultant Allison Andrews — returning from Episode 2 — to walk through the reality behind the industry's favorite euphemism. Four massive music datasets. Twelve million tracks. A million-million copies made in GPU memory. And the asymmetry at the heart of it: Zebulon Sheridan's stolen songs were scrubbed from every platform, but AI models trained on copyrighted music are still live, still generating, still profitable.

References from this episode:

• Is Your Work in a Training Dataset? — Search tools for images, books, music, and web content
• Documenting Copyright Infringement — Evidence preservation checklist
• Have I Been Trained — Search LAION-5B for your images
• The Atlantic AI Watchdog — Search music datasets by artist name
• Glaze & Nightshade — Data poisoning tools from UChicago

Next episode: The Law — DMCA takedowns, class actions, and what the platforms actually have to do.

Featuring Heavy Comforter music throughout. HeavyComforter.com — the real thing.

With JS going full Paul Harvey on the 4,854-word Finnegans Wake recitation — what the echo remembered, what it forgot, and why the forgetting might be the point. One question that won't resolve: is an echo art?

Featuring Heavy Comforter music throughout. HeavyComforter.com — the real thing.

For the full resource — data center map, community opposition tracker, references — visit beaumontandsheridan.com/resources/data-centers-the-internets-body. Featuring Heavy Comforter music throughout. HeavyComforter.com — the real thing.

John Shiloh lays out the Zebulon origin story: a fraudulent imposter singing Heavy Comforter songs at slowed-down speed, uploaded to every major streaming service by an unknown content farmer. The record did well. Real money was made. No legal action was taken — a weird little case nobody's heard of, but John Shiloh's law office is on it.

Zebulon punctuates with a lyrical meltdown — \"I am here by the grace of nothing / I'm depressed, but I guess that's something\" — before John Shiloh restores order. The legal disclaimer: you can't take legal advice from a podcast, no matter how good the lawyer sounds. Talk to a real one.

Featuring Heavy Comforter music throughout. HeavyComforter.com — the real thing. Next episode: data centers — the internet's body. Resource page at beaumontandsheridan.com.

Ad: Good Eats Cafe, East 23rd Street, Chattanooga TN. 24 hours. Meat and three. Red chairs. Bob Seger on the jukebox. No record in any training corpus — lost to time.