Alex Karp sat down on CNBC's "Squawk Box" on July 1 and said what every enterprise CEO has been thinking in private. "Something has gone completely wrong," the Palantir CEO told the hosts. "The basic view among enterprises in this country is: I'm going to chillax and waste my time with tokens, I'm going to get no value, and they're going to get my IP."
The clip went everywhere. Social media. Boardrooms. Group chats between founders. Most of the coverage framed it as an attack on large language models. It was not. Karp was not arguing that the models are bad. He was arguing that the way they are sold is a trap — a trap designed to extract the most valuable thing a company owns: its institutional intelligence, its processes, its competitive edge. He called it your "alpha." And he said the AI labs are taking it.
Palantir has published a nine-point manifesto on AI sovereignty. Point two: "Data retention is your treasure. Transfer it at your own peril. Your ability to win is dictated by your ability to recognize and use your unique edges, and you keep winning by compounding the underlying data to generate new insights. Transferring that data hands over access to your pre-existing winning plays and yields the means of production for new ones."
Point four: "Controlling your weights is controlling your fate. Weights are the distilled form of hard-won, accumulated institutional knowledge. If you let others control your weights, you are allowing them to migrate the alpha of your business to theirs."
That language should chill every PR firm principal reading this.
The opt-out is a false wall
Most PR firms believe they have solved this problem. They checked the box that says their prompts will not be used for training. They read the terms of service. They feel protected.
They are not.
The frontier AI labs do not need your training data to learn your business model. Every time your team engages with their API, the platform sees the structure of how you work. The sequencing of your research. The way you frame a pitch. Your media training process. The way you manage a crisis. The media targeting logic that connects a client's message to a specific reporter at a specific outlet. The workflow itself is the intelligence — and the platform is watching it execute in real time, thousands of times a day, across every firm that uses it.
Opting out of training is not the same as opting out of observation. The models are not learning from a static dataset of your old prompts. They are learning from the living patterns of how an entire industry operates. Your pitch structures, your editorial instincts, your angles, your sequencing — it all flows through infrastructure you do not own, governed by terms you did not write, controlled by companies whose economic incentive is to productize exactly what they are seeing.
And here is the part that should concern the best firms most: the models learn from everyone. The shops doing exceptional work and the shops producing mediocre output are contributing to the same pool. The eventual product is not a replica of the industry’s best practices. It is an average — a composite built from every operator, good and bad, that has ever run a workflow through the platform. The craft knowledge that took your firm decades to build gets diluted into a generic capability and sold back to every competitor at the same subscription price.
The best firms lose the most. Their alpha gets blended into a product that makes the worst firms slightly better and the best firms significantly less differentiated. That’s not a bug in the system. It’s the business model.
Exhibit A: what Anthropic did to Figma
If you want to understand what Karp is warning about, look at what happened between Anthropic and Figma.
Figma holds 80% to 90% market share in UI and UX design. Anthropic partnered with them. Figma built integrations around Claude. In February, they launched a joint feature called "Code to Canvas" that connected Claude Code directly to Figma's design workflow. The companies were collaborators.
Then in April, Anthropic launched Claude Design — a standalone product that generates interactive prototypes, slide decks, and marketing collateral from conversational prompts. A direct competitor to Figma's core business. Anthropic had recruited Figma as a co-announcement partner for the launch. Figma allegedly completed negotiations and was prepared to announce alongside them.
Anthropic launched without them.
Figma's CEO, Dylan Field, publicly said Anthropic had not been "consistently candid" in its communications. The Information reported that Anthropic had told partners Claude Design would be "fairly basic" — no advanced collaboration, no ability to edit AI-generated designs. Features that competed directly with Figma and Canva. Then the product shipped with seemingly exactly those features.
A senior Anthropic executive resigned from Figma's board three days before the launch.
This is what it might look like when a platform partner studies your workflows, absorbs your domain expertise, and then builds the product that replaces you.
This is coming for PR*
Every PR firm using a frontier model is feeding its operational intelligence into that model's ecosystem. Not through a training checkbox. Through use.
How far are we from Claude Media? From a platform that packages what it has observed across thousands of PR professionals and sells it as a product? From a tool that takes the accumulated craft knowledge of an entire industry — the pitch patterns, the editorial instincts, the media targeting strategies — subjects it to the tragedy of the commons, and offers it to anyone with a subscription?
Ask Figma how quickly a partner becomes a competitor.
Satya Nadella reinforced this warning weeks before Karp's interview. The Microsoft CEO wrote that entire industries could "find their knowledge commoditized right out from underneath them." He said the real opportunity is not in picking the best model but in building a learning loop where human expertise and AI capability compound together, inside infrastructure the company owns.
Karp told CEOs to ask two questions: "Are you keeping the data? Are you (frontier companies) going to enter our business?"
PR firm principals should be asking those questions today. And they should add a fourth: When the platform has learned enough from watching my industry work, will I lose my competitive edge?
If you cannot answer that question, you already know the answer.
---
Greg Matusky is founder and CEO of Gregory, a financial services PR firm ranked among the largest in the country.