<p>In the US, an important court case against Anthropic (the creator of Claude) has taken place. The judge ruled: training AI on books without the authors' permission is permissible if the books are obtained legally. However, creating an internal library from pirated copies is a direct violation of copyright.</p>
<p>Anthropic initially downloaded millions of books from pirated sites like LibGen, Books3, and Pirate Library Mirror. Then the company purchased printed books, digitized them, and also used them for AI training. All of this was done to build a central library that they planned to keep "forever." </p>
<p>The court recognized that:</p>
<ul>
<li>creating a "perpetual" library from pirated books is a violation of the law;</li>
<li>training AI on these books is fair use if the copies were purchased or obtained legally;</li>
<li>the issue of compensation for copyright infringement is still unresolved and will be addressed in a separate proceeding.</li>
</ul>
<p>For the industry, this means: if AI companies do not pirate data, they can continue to train LLMs on purchased or open-access books. The legality of scraping from open websites has not yet been considered separately.</p>
<p>This is just the beginning: there are many similar cases, and the final position of the courts has not yet been determined. But the ruling is a clear win for AI companies and bad news for authors.</p>
<p>π Source: <a href="https://storage.courtlistener.com/recap/gov.uscourts.cand.434709/gov.uscourts.cand.434709.231.0_2.pdf">https://storage.courtlistener.com/recap/gov.uscourts.cand.434709/gov.uscourts.cand.434709.231.0_2.pdf</a></p>;
<p>#ai #copyright #anthropic #claude #law</p>
<p>βββββββββ<br>ΠΡΡΠ»ΠΈ Π Π²Π°ΡΠ΅Π²Π°<br>βββββββββ</p>
