The IP Framing Problem in AI
Most AI moat arguments are intellectual-property arguments in disguise, and the IP mostly isn't there. Five parts on where defensibility actually lives.
The industry talks about data moats, prompt libraries and agent IP. Five parts on why most of that framing does not survive contact with either the law or the market.
Your data is not a moat in the way the pitch decks claim. The prompt-protection playbook rests on a layer the Copyright Office already concluded does not confer authorship. Agents extended the same framing onto a layer where it makes even less sense, and produced some of the most expensive prompt wrappers in business history.
The closing part looks at the lawsuits everyone is watching — NYT v. OpenAI, Bartz v. Anthropic, Thomson Reuters v. Ross — and argues the market has been telling a different story the whole time.