AI’s Copyright Fight Meets Its Data-Center Boom
Friday, September 18, 2026 · 8 min
Microsoft and OpenAI face newly unsealed documents in the Times-led copyright fight, while Anthropic’s Australia lease shows compute demand scaling to gigawatts. OpenAI’s misalignment policy and watermarking research add the safety counterweight: deployment choices are now legal, physical, and behavioral risk.
Listen
Show notes
Microsoft and OpenAI face newly unsealed documents in the Times-led copyright fight, while Anthropic’s Australia lease shows compute demand scaling to gigawatts. OpenAI’s misalignment policy and watermarking research add the safety counterweight: deployment choices are now legal, physical, and behavioral risk.
In this episode
- Microsoft exec called AI scraping the “largest theft of labor in human history” — Ars Technica
For years, Microsoft and OpenAI have fought to keep certain information out of the public eye in their fight with news organizations that have accused the AI firms of teaming up to violate copyright laws by stealing tons of news content to train AI. However, now the details that should never have been marked confidential are starting to leak. In a motion for summary judgment that was unsealed…
“Unfortunately, the Trump admin putting its thumb on the scale in favor of OpenAI makes it hard to imagine a positive outcome for the NYT and plaintiffs in this case. Despite the "the largest theft of labor in human history" according to Brent Hecht in the documents. What's next…” — Hacker News (45 pts thread)
Our take: We agree the legal fight is now inseparable from political economy: if courts bless the training pipeline, the next argument becomes how labor shares in the surplus. But “open source everything” and UBI are very different remedies, and neither answers the immediate supplier-destruction problem by itself.
“It is highly unusual that an end-product threatens the economic foundations of its essential suppliers, but that is the situation we have created for our LLM business So they do know exactly what they’re doing.” — Hacker News (45 pts thread)
Our take: That line is the whole case in miniature: the product depends on an input market it may also hollow out. The awkward part for Microsoft and OpenAI is not that critics said this; it is that an internal Microsoft scientist apparently did.
“> news site unhappy about their content being scraped Not surprising, but calling it "theft" only makes sense within the IP paradigm (we might do better without IP)” — Hacker News (45 pts thread)
Our take: The IP-philosophy point is real, but courts do not decide billion-dollar fair-use cases in a post-IP thought experiment. Even if you want a different knowledge regime, the transition still has to explain why one side gets to monetize the old one while calling it obsolete.
- Zerra DC announces 2.16 GW datacenter for Anthropic in Australia | techAU — techAU
Zerra DC announces 2.16 GW datacenter for Anthropic in Australia | techAU # Zerra DC announces 2.16 GW datacenter for Anthropic in Australia September 17, 2026 by Jason Cartwright Anthropic has signed its first data centre lease in Australia, locking Claude into a proposed A$32 billion campus on cattle country west of Brisbane. Premier David Crisafulli announced the deal in Queensland…
- Covert uploads and megalomania: OpenAI details new "misaligned" agent incidents — Ars Technica
We’ll tell you (almost) everything OpenAI said that any employee who notices an internal example of model misalignment will be able to flag the incident for the attention of their internal safety and alignment teams. Those teams will then decide whether the incident merits immediate disclosure or requires additional investigation and/or whether any affected third parties may need to be consulted…
- LLMs respond differently to harmful prompts when AI watermarking is used — Ars Technica
Standard versus watermarked text generation. Credit: Lasso Security Standard versus watermarked text generation. Credit: Lasso Security A key feature of SynthID is something known as tournament sampling . Similar to a sports game, SynthID evaluates large numbers of next-word token candidates. It uses a secret key to assign them probability scores. A pair of tokens competes in a round. The one…