In the News

According to MSN, GenAI developer Anthropic has agreed to pay at least $1.5 billion to settle a class action lawsuit alleging the firm pirated thousands of copyrighted works to train its large language model Claude.

In June, U.S. District Judge William Alsup ruled that Anthropic was within its “fair use” rights to use legally acquired works to train its LLM, because the platform transformed the material into “something new.” However, Alsup also found Anthropic liable for its improper use of millions of pirated ebooks.

According to attorneys for both sides, a preliminary settlement has been reached. Details have not been disclosed but reports suggest a $1.5 billion settlement covering about 500,000 pirated works is on the line. Anthropic would also be required to destroy any pirated materials that had been used to train its LLM, essentially closing the barn door after the horse has been stolen and eaten.

Anthropic’s latest round of fundraising brought in $13 billion. The company is valued at $183 billion, three times its reported value as recently as last March, suggesting investors were not scared off by the lawsuit, but rather consider it a cost of doing business.

How the settlement will influence other pending litigation remains to be seen.

Writers recently won a partial victory in a class action lawsuit against generative AI firm Anthropic.

The lawsuit alleged that Anthropic trained its language learning model, Claude (Claude?), using millions of copyrighted works scraped from pirate sites.

In a recent ruling, US District Court Judge William Alsup ruled that Anthropic’s use of purchased or otherwise legally acquired material to train its AI was fair use, even though the authors were never contacted and never consented to Anthropic using their work in this way.

However, the judge differentiated between legally acquired work and work that had been stolen, even though Anthropic wasn’t responsible for uploading the original files to the pirate site.

Plaintiffs established a certified class of injured parties, comprised of rights holders whose registered books had been pirated and then later used by Anthropic. Writers concerned about the possible misuse of their work can use this database on The Atlantic website to see if their work was uploaded to LibGen, one of the sites scraped by Anthropic.

If you find any of your work in that database, you can submit your information to a website created by plaintiffs’ attorneys to gather data on injured parties in the class. The deadline for submitting information was August 15, but that deadline may be extended. If you think you’ve suffered damages, keep trying. Joining the class will ensure you receive additional notices about the case, as well as the chance to remain part of the class and potentially receive a portion of any recovery.

And the recovery in this case could be substantial.

Statutory damages range from $750 US to $150,000 per infringed work. If the infringement is deemed willful – ie: if Anthropic knew that it used stolen work to train its AI – the court would likely look for damages higher than the minimum.

Breaking down the math: If the court finds that 100,000 copyrighted works were infringed, damages could amount to billions of US dollars. In an extreme case, if the court finds that Anthropic pirated some 6 million copyrights works, damages would exceed $1 trillion US. Even the minimum of $750 per infringement of only 100,000 would equal $75 million.

In any case, a ruling for the plaintiffs would likely be the end of Anthropic, while the precedent could haunt the cases pending against generative AI giants like ChatGPT and Midjourney. With powerhouses like the New York Times and Disney behind those, those cases won’t be disappearing anytime soon.

Know anyone who’d like my blog? Please forward today’s post! I’d love to hear from them.

Need more content? Join my mailing list!

If you want to ride in a deathtrap but are afraid of the water, I have a 20-year old Jeep with a recall I’ve been ignoring. I’ll take you wherever you want to go for only $25,000 per seat. You may not see the wreck of the Titanic, but you can roll down a window for air anytime you want.

Inquire within.

By all accounts, San Francisco’s downtown retail environment is in full collapse. Last week, the Westfield San Francisco Centre announced its immediate closure, just weeks after Nordstrom said it would leave the city center. They weren’t the first major outlets to abandon the city, and there have been more since, including GAP, Old Navy, and AT&T.

So what to do with all that empty space? Will other retailers want to take over a spot that’s lost a third of its business? Will the mall become a homeless squat?

I have a better idea.

YouTube buys the mall and converts the stores into high-end hybrid studio/apartments to rent out to vloggers and influencers. They broadcast live from their apartments and people can watch them during the mall’s operating hours. Instead of shopping or engaging in any activity, the public can simply mill about and stare. Boom! Tourism! Keep the glass storefronts so influencers can live their entire lives on display. Boom! Happy attention whores! The audience can tip with credits for the food court. Boom! High cholesterol, fat, and sodium!

I think this one has merit.