Can AI Be Trained Ethically? New Dataset Shows It’s Possible

A team of researchers has shown that it’s possible to build large AI datasets entirely from ethical sources, namely 130,000 English-language books from the Library of Congress—almost twice the size of Project Gutenberg’s collection. Their project adds to recent open-source efforts like Hugging Face’s FineWeb, which aim to make AI training more transparent and responsible. While experts say this careful approach may not be big enough to power today’s largest AI models, they hope it will encourage companies to be more open about what data they use. CLICK HERE FOR FULL ARTICLE

Video Game Strike Ends Addressing AI Voice-over Concerns

After nearly a year on the picket line, voice and performance-capture actors in the video game industry have reached a landmark deal with publishers — a moment that firmly places gaming companies in the same labor league as Hollywood’s biggest studios.

For decades, video game giants were treated as tech companies, striking union agreements tailored to their unique medium. Under the new Interactive Media Agreement, they’ve joined TV and film studios in accepting union terms on key issues like AI use, informed consent, and fair compensation.

The contract delivers wage increases and AI protections. Negotiating committee chair Sarah Elmaleh highlighted one especially critical element: language preventing “Generative AI scabs” from replacing an actor’s labor with a simulated version of themselves during a strike.

“It may sound wonky,” Elmaleh said, “but every worker deserves to know that if they withhold their labor, they can’t be replaced by a version of themselves.” Click Here for More Information

.

The Copyright Quagmire...where are we headed?

In a July 23, 2025 speech unveiling his “AI Action Plan,” President Trump stirred controversy by dismissing the idea that AI developers should pay for every article, book, or piece of content used as training data. He argued that such a requirement is simply “not doable” and would make it impossible to build successful AI systems. See Deadline Article. His underlying comparison drew on how humans learn—“When a person reads a book or an article, you've gained great knowledge. That does not mean you're violating copyright laws or have to make deals with every content provider” See Patently-O

and Politico

What This Means for Copyright Law

• Fair Use vs Fair Training? Trump’s remarks align with a growing argument that AI training should qualify as fair use—especially when it involves transforming and learning from content, not copying it outright.

  
  

• Licensing Chaos: Requiring AI developers to negotiate licenses for each work used could create a logistical nightmare: millions of contracts, lengthy negotiations, and soaring costs.

  
  

• But What About the Rights of the Copyright Owner? Publishers and rights holders warn that eroding copyright protections will chill creative content. They argue that without proper compensation and incentives, original creators will be undercut and undervalued.

Possible Solutions

There are some that suggest a middle-ground solution such as compulsory licensing pools, which bundle rights into one standardized system for payment and access. This may simplify the licensing process while still respecting the intellectual property rights of the content owner. patentlyo.com.

Senators Josh Hawley (R-Mo.) and Richard Blumenthal (D-Conn.) have introduced a bill that would prohibit AI companies from training on copyrighted works without permission, empowering creators to sue for misuse. See Publishers WeeklyArticle. Amazon recently entered into a $20+ million per year deal to license content from The New York Times suggesting that AI modelers are willing to honor copyright law when they can license a large body of works from a single point of access. The Wall Street Journal