Frontier AI Labs with Wil Waldon
Aug 4, 2026 · 12 min · 10 segments
**$1.5 billion settlement** in the copyright lawsuit between authors and the AI company **Anthropic**. The litigation centered on allegations that the developer utilized **pirated…
Well, part of the deal Anthropic agreed to is destroying the allegedly pirated data sets they used.
Which is such a weird concept because once an AI model has ingested and tokenized half a million books, deleting the original text files doesn't erase the patterns and the connections the model already learned.
But internally, we have emails from Anthropic CEO Dario Amadei revealing the company consciously decided to bypass formal licensing negotiations for their initial models anyway.
They just ignored the licensing part entirely.
The executive leadership actually framed those negotiations as a legal practice and business slog that would just delay their commercial roadmap.
So they just skip it to prioritize speed.
Where returning or destroying the stolen goods supposedly fixes the problem.
But machine learning just does not operate that way.
You can't untrain the weights.
And because of that friction, AI companies are using a parallel strategy now, which is buying physical out of print books and destructively scanning them.
Wait, destructively scanning them, like actually destroying the physical books?
Yeah, literally running them through paper shredders equipped with high speed scanners to get pristine human written text that is entirely free from Internet generated AI pollution.
I mean, the irony of literally destroying physical knowledge just to feed an algorithm to build digital knowledge.
It just creates this completely bizarre, ethical gray area.
Well, part of the deal Anthropic agreed to is destroying the allegedly pirated data sets they used.
Which is such a weird concept because once an AI model has ingested and tokenized half a million books, deleting the original text files doesn't erase the patterns and the connections the model already learned.
But internally, we have emails from Anthropic CEO Dario Amadei revealing the company consciously decided to bypass formal licensing negotiations for their initial models anyway.
They just ignored the licensing part entirely.
The executive leadership actually framed those negotiations as a legal practice and business slog that would just delay their commercial roadmap.
So they just skip it to prioritize speed.
Where returning or destroying the stolen goods supposedly fixes the problem.
But machine learning just does not operate that way.
You can't untrain the weights.
And because of that friction, AI companies are using a parallel strategy now, which is buying physical out of print books and destructively scanning them.
Wait, destructively scanning them, like actually destroying the physical books?
Yeah, literally running them through paper shredders equipped with high speed scanners to get pristine human written text that is entirely free from Internet generated AI pollution.
I mean, the irony of literally destroying physical knowledge just to feed an algorithm to build digital knowledge.
It just creates this completely bizarre, ethical gray area.
The rest of this transcript — segmented and speaker-labeled, so you land on the exact moment something was said
Search every transcript — by keyword, by phrase, or by meaning, across every show Radar indexes
Trends — what is surging across podcasts, measured against its own baseline
Alerts — when a name you follow appears in a newly indexed episode
No account is needed to search Radar.