Snap Settles Teen Addiction Case as AI Copyright Fight Widens

Snap reached a settlement on 2 November in a lawsuit brought by two teenagers, identified in court documents as K.D.B. and S.J. They said the platform was designed to be addictive and contributed to depression, anxiety and eating disorders. The terms were not disclosed. Jury selection had begun the previous week and the trial was due this month, according to Engadget.

The stakes were specific. Spiegel was expected to take the stand. That would have been the first time the Snapchat creator appeared in court in a social media addiction trial. Snap was the only remaining defendant after Meta settled the previous week. The plaintiffs had already settled with Meta, YouTube and TikTok.

“The Parties are pleased to have been able to resolve this matter in an amicable manner,” a Snap spokesperson said in a statement to TechCrunch. The company added that it has “built safeguards, launched safety tutorials, and partnered with experts.” It said it remains committed to strengthening “the safety, privacy, and well-being of our community.”

A pattern of settlements, and one jury verdict

The settlement fits a pattern. Earlier this year, Meta, YouTube, Snap and TikTok faced another case. A 20-year-old woman said the apps harmed her as a child. Snap and TikTok settled before trial. A jury then ruled against Meta and YouTube and ordered them to pay $6 million in damages, Engadget reported.

Emily Jeffcott, an attorney for K.D.B., said in a statement: “We are proud to have achieved justice for our bellwether clients so far, including K.D.B. and S.J. Our ultimate goal is to hold these social media companies accountable for how their actions have impacted millions of children across the country, and these settlements allow us to keep building toward that goal.”

Both Meta and Snap still face numerous other lawsuits over addiction and other alleged safety issues. The plaintiffs said they started using YouTube as young as age 3 and other platforms at around age 9. They blamed features such as autoplay and infinite scrolling for compulsive habits, according to TechCrunch.

The same legal machinery, now aimed at AI

What connects this settlement to the AI copyright wave is the method. It relies on mass litigation, bellwether cases, and defendants settling rather than testing a legal theory in front of a jury. The docket against AI companies is filling up on similar logic. Publishers and musicians argue that training and output amount to infringement.

Recent headlines show the scale. US newspapers sued Brave Software and Canadian publishers sued Anthropic, according to Press Gazette. The Orlando Sentinel reported that Brave “plundered” its content and resold it to AI companies. An EU legal adviser said a Google AI chatbot may break press copyright rules, ETLegalWorld.com reported on 3 November.

Music is moving in parallel. Anthropic faces a $41 million copyright lawsuit over 274 Elvis Presley songs used to train AI models, according to a report dated 26 October. Anthropic and Suno were also sued over alleged training on recordings by Madonna, Johnny Cash and Joni Mitchell. Reports on 28 October put music copyright lawsuits at $338 million. Sony Music Entertainment filed a new lawsuit against AI startup Udio in July.

None of these numbers come from court filings. They come from news reports, and should be read as reported figures, not adjudicated ones.

What the infrastructure story says about scale

The reason the legal wave keeps growing is visible in the buildout. Google Cloud and NVIDIA announced general availability of NVIDIA Vera Rubin infrastructure on 21 October. A5X instances are coming next month, with planned expansion to Europe. Google Cloud plans to scale its Vera Rubin-based infrastructure to as many as 960,000 Rubin GPUs across multiple sites, according to the NVIDIA blog.

Ineffable Intelligence, a London company founded by AlphaGo creator David Silver, was the first customer to access A5X. “We’re building superlearners that continually improve through their own experience,” Silver said in the announcement. “That demands both substantial compute and a tight connection between generating experience and training on it.”

WPP demonstrated a creative agent architecture on A5X instances using NVIDIA Nemotron models and OpenShell. In its initial tests, workloads that typically required a multi-node cluster ran on a single A5X instance, according to the same post. The more capability that gets packaged into single instances, the more content gets generated. The more surface area copyright holders see.

China’s inference milestone

Usage data points the same way. A report released on Saturday by the China Centre for Information Industry Development said China’s average daily token calls exceeded 175 trillion. That is more than 1,000 times the level in early 2024, the South China Morning Post reported on 2 November. The centre operates under China’s Ministry of Industry and Information Technology.

For the first time, the report says, inference data generated by AI systems in China has surpassed the data used to train those systems. Inference means running trained models to answer requests. More inference means more output, and output is where copyright claims bite.

Neither report says whether the CCID report addresses licensing or infringement. It presents the token figure as a measure of adoption.

Power, land and the physical bill

Behind the compute is electricity, and the gap between data center timelines and grid timelines is widening. An Ember analysis published in October 2026 found that in Ireland, data centers already account for about 19% of peak demand. That is 1.1 GW of a 6 GW peak in January 2025. Ember projects that share could rise to 24% by 2030, with data center demand reaching 1.6 GW against a 6.5 GW system peak, pv magazine reported on 3 November.

A large data center can need 200 MW to 500 MW and be built in 18 to 24 months. Grid expansions can take four to 10 years. Load shifting could cut Ireland’s peak by about 100 MW by 2030, or 1.5%. Behind-the-meter generation and storage could provide nearly five times more relief, with potential reductions up to 2.7 GW in Texas, according to the same analysis.

Storage is being added at the source. As of September 2026, over 16% of new utility-scale solar capacity added this year is co-located with operable battery storage. That compares with 2% in 2017, according to EIA data reported by CleanTechnica. Plans for the last three months of the year put the share at 19%, rising above 30% in plans for 2027 and 2028.

Those figures describe the physical cost of the AI buildout, not its legal cost. Both are rising at once.

Retail, rights and the money around the fight

On 3 November the European Commission approved JD.com’s takeover of MediaMarktSaturn with conditions, heise online reported. The offer values Ceconomy, the parent of MediaMarkt and Saturn, at around €2.2 billion. Brussels found that JD.com had received Chinese subsidies through preferential financing, tax breaks and grants. It required that intra-group transactions be at market terms. It also required that JD.com offer warehousing, logistics and technology services to other retailers on the same terms as Ceconomy. The commitments bind the companies for ten years.

This deal is not connected to AI copyright, and it should not be read as part of it. It is included here because it is the newest dated item and shows the regulatory mood around large platforms.

Sports rights show where media money is going. Inter and AC Milan unveiled a naming rights deal with Generali for a new 71,500-seater stadium. SportsPro reported on 30 October that the deal is worth €650 million (US$725 million) over 20 years. The new venue is expected to open in the 2031/32 season, with work to begin in 2028.

At Sportel in Monaco, IMG’s Hillary Mandel said properties cannot depend entirely on their media deal: “The dollars are incredibly important, but the ecosystem that sits around that… you’ve got to own it yourself,” SportsPro reported on 26 October. Mediapro, rebranding as Imagina, said traditional rights models “don’t work anymore,” according to SportsPro on 20 October.

What to watch

Two things are unresolved. First, the Snap settlement terms are undisclosed, so there is no number to compare with the $6 million jury award against Meta and YouTube. Second, the AI copyright cases cited here are at filing or opinion stage. None has produced a final judgment.

The pattern from the addiction litigation is that defendants settle before a bellwether reaches a verdict. If AI companies follow the same path, the reported figures will keep arriving through settlements rather than rulings. The public record will stay thin.