Zuckerberg 'Personally Authorized and Encouraged' Meta's Copyright Infringement

Alleged conduct and scale

• Commenters focus on claims that millions of copyrighted books and articles were torrented (tens of terabytes) and used to train large language models, including seeding torrents and stripping copyright metadata.
• Some compare potential damages to a prior Anthropic settlement (billions of dollars; thousands per infringed work). Others note Meta’s huge cash and market value, arguing even multi‑billion fines might be a “slap on the wrist.”

Punishment, inequality, and corporate liability

• Strong sentiment that ordinary people and small-time “pirates” faced ruinous lawsuits and even prison, while large tech firms and executives expect at most monetary settlements.
• Many call for personal consequences (including jail) for executives who directed infringement, not just corporate fines, and criticize limited liability and current corporate personhood.
• Others argue criminal penalties for copyright are themselves unjust, but still insist that, until laws change, they should be applied equally to powerful actors.

Legality of AI training vs piracy

• One camp: training on copyrighted works is “like reading,” inherently transformative and fair use; the real legal problem is unlicensed acquisition (torrenting, ignoring robots.txt, piracy).
• Another camp: training necessarily involves copying; scale and commercial substitution matter, and models that regurgitate or function as market substitutes cross into infringement.
• Disagreement over whether current copyright law distinguishes humans vs machines, and whether scale (millions of works, industrial use) should change the analysis.

Views on copyright itself

• Several participants are anti‑copyright or “copyright abolitionist,” but still want laws enforced uniformly against big AI companies, partly as poetic justice for past harsh enforcement against individuals.
• Others favor reform: stronger protection for individual creators, weaker power for large rights-holders, and clearer rules for AI training and sampling‑like uses.

Enforcement mechanisms and precedent

• Some suggest RICO or criminal copyright statutes could apply if executives knowingly directed mass infringement. Others think the legal terrain around AI and fair use is still “clear as mud.”
• Comparisons recur to cases against file-sharing site operators, student MP3 downloaders, and high-profile prosecutions over academic journal downloads, highlighting perceived double standards.

Operational and ecosystem concerns

• Multiple reports of Meta’s crawlers aggressively scraping sites and ignoring robots.txt, prompting ASN‑level blocking.
• Open release of model weights is seen as making ongoing royalties for inference practically unenforceable, even if training is later found infringing.