SAN FRANCISCO— Newly unsealed court transcripts from a landmark intellectual property lawsuit have exposed a deep, existential dread over AI scraping within the tech sector. The documents reveal a tense deposition of a prominent software executive. Under oath, the executive struggled to defend the legal basis of training large language models on copyrighted data. Legal experts say the testimony confirms that tech leaders view AI scraping as an unsustainable system that will eventually destroy its own information sources.
PROCEEDINGS OF THE U.S. DISTRICT COURT FOR THE NORTHERN DISTRICT OF CALIFORNIA
CASE NO. 24-CV-89021
TESTIMONY OF DR. LAWRENCE VANCE, CHIEF DATA ACQUISITION OFFICER
MR. COOPER (COUNSEL FOR PLAINTIFFS): Dr. Vance, let us discuss how your company trains its algorithms. You use a process called AI scraping, correct?
DR. VANCE: Yes. We crawl the open internet. We ingest public text to teach our models how humans communicate.
MR. COOPER: And by 'ingest,' you mean you copy entire databases without permission or compensation?
DR. VANCE: We prefer the term 'automated learning.' It is similar to a human reading a book at a library.
MR. COOPER: Does a human memorize three billion books in four seconds and then sell access to those memories?
DR. VANCE: That is a matter of scale, not substance.
MR. COOPER: Let the record show the witness is sweating profusely. Dr. Vance, I am showing you an internal email you sent on January 12. In it, you describe AI scraping as 'the largest heist since the invention of the wheel.' Do you recall writing that?
DR. VANCE: I wrote that during a late-night brainstorming session. It was a metaphor. We were feeling dramatic.
MR. COOPER: You also wrote, 'If we do this for two more years, there will be no journalists left to steal from, and our bots will start eating their own digital waste.' Is that also a metaphor?
DR. VANCE: That refers to our internal concerns regarding data cannibalism. It is a technical term.
MR. COOPER: It sounds like you are describing a doom loop. You destroy the writers, which destroys your training data, which destroys your product.
DR. VANCE: Yes, but our quarterly revenue is up forty percent. We expected to be on Mars before the internet ran out of words.
MR. COOPER: Dr. Vance, is your entire business model just a giant digital woodchipper?
DR. VANCE: (Long pause) The woodchipper is highly efficient. It creates a lot of value for our shareholders.
MR. COOPER: No further questions.
At press time, the defense team filed a motion to replace Dr. Vance with a friendly generative chatbot that has been programmed to plead the Fifth Amendment in six million different languages.
