The New York Times has taken its fight against Perplexity to a new level, filing a lawsuit in federal court that accuses the fast-growing AI startup of systematically lifting, reproducing, and monetizing the newspaper’s journalism without permission.

The complaint, submitted Friday in the Southern District of New York, alleges that Perplexity “unlawfully crawls, scrapes, copies, and distributes” Times content — often in forms that the Times says are either verbatim reproductions or so close that they function as substitutes for the original work. The lawsuit says the Times repeatedly warned Perplexity to stop, issuing cease-and-desist letters last year and again in July, but the company allegedly continued the practice.

The accusation arrives after months of scrutiny from news outlets whose paywalled reporting has been turning up inside Perplexity’s summaries. Investigations by Forbes and Wired earlier this year found that the startup’s systems had been bypassing standard protections — including the industry-accepted robots.txt file — to harvest material behind paywalls.

The Times is making the same claim, arguing that Perplexity’s crawlers “intentionally ignored or evaded technical content protection measures” that publishers use to control which parts of their sites automated tools can access.

The paper argues that the consequences are direct and measurable: if Perplexity delivers Times reporting to users in ready-made form, the user no longer needs to subscribe, click through, or visit the Times website. That, the lawsuit says, undercuts the revenue streams the outlet depends on — subscription, advertising, licensing, and affiliate income — all while Perplexity positions the appropriated content inside a for-profit AI product.

Perplexity, founded in 2022, markets itself as an “answer engine” built on generative AI. Its pitch is simple: instead of searching, you ask questions, and the system produces a consolidated answer drawn from the web. The model has gained traction quickly, attracting investors and positioning itself as a challenger to ChatGPT and Google’s AI search experiments. That rise has also meant increasing friction with publishers who believe Perplexity is building a business on other people’s work.

The lawsuit lands amid a broader wave of legal pressure on the startup ecosystem. One day before the Times filed its case, the Chicago Tribune brought its own copyright lawsuit against Perplexity, also accusing the company of bypassing paywalls and repackaging its content. More broadly, the Times itself is no stranger to legal battles with AI firms: it sued OpenAI in December 2023 over similar claims, before eventually reaching a licensing deal that allows Amazon to bring Times content to products like Alexa.

The legal landscape around AI training and output remains unsettled. Major AI firms argue that scraping publicly available content is protected under fair use, and that generated summaries are distinct creative outputs. News publishers counter that AI systems powered by their articles are siphoning off their audience and financially undermining the journalism that AI companies rely on.

Perplexity has attempted in recent months to show publishers it wants a cooperative relationship, launching a revenue-sharing program last year and later expanding it to include its Comet web browser in August. But the Times says the startup’s current practices still amount to unauthorized copying.

“By copying The Times’s copyrighted content and creating substitutive output derived from its works, obviating the need for users to visit The Times’s website or purchase its newspaper, Perplexity is misappropriating substantial subscription, advertising, licensing, and affiliate revenue opportunities that belong rightfully and exclusively to The Times,” the complaint states.

Perplexity pushed back in a statement sent to The Verge. Spokesperson Jesse Dwyer said: “Publishers have been suing new tech companies for a hundred years, starting with radio, TV, the internet, social media and now AI. Fortunately it’s never worked, or we’d all be talking about this by telegraph.”

The Times is seeking damages and a permanent injunction that would bar Perplexity from what it describes as unlawful scraping and redistribution of its journalism. What the federal court decides could set one of the most consequential precedents yet in the standoff between generative AI companies and the news industry — a conflict that now stretches from OpenAI to Google, xAI, and a long list of AI upstarts whose systems rely on the work of reporters they neither employ nor pay.

Meta has struck new commercial agreements with major news publishers to feed its Meta AI chatbot with real-time global, entertainment, and breaking news, marking a sharp pivot from the company’s retreat from news distribution in recent years.

The tech giant said on Friday that users who ask Meta AI news-related questions will now get responses that surface information and links drawn from partner outlets. Those links point directly back to publishers’ websites, which Meta says will help those partners reach new audiences.

The initial group includes CNN, Fox News, Fox Sports, Le Monde Group, the People Inc. portfolio of media brands, The Daily Caller, The Washington Examiner, and USA Today. Meta plans to add more partners over time.

The shift reverses years of deliberate scaling-back: Meta killed Facebook’s News tab in 2024 and stopped compensating publishers as far back as 2022. Now it is paying again — not to revive news feeds inside its apps, but to supercharge Meta AI’s accuracy, speed, and usefulness by giving it real-time access to professional reporting.

“We’re committed to making Meta AI more responsive, accurate, and balanced,” the company said.

It noted that rapid live-event coverage is especially hard for current AI systems to handle, and argued that onboarding multiple news sources would help balance viewpoints and deliver more timely information.

The move comes amid rising pressure for Meta to stay competitive in the AI race. Llama 4’s controversial early-year rollout drew complaints of weak performance, at a time when rivals are pushing out increasingly powerful models. Meta AI is already available in more than 200 countries across Facebook, Instagram, WhatsApp, Messenger, and a standalone app, but user adoption remains a key metric.

The Great Scramble for Quality Data

Meta’s new deals land in a world where the entire AI industry is straining under a “content crunch.” Large language models are trained on enormous quantities of text, much of it scraped freely from the open internet, but several of the highest-value information sources have begun to push back.

That backlash is driven by two forces:

1. Big Tech’s model-training habits impose steep costs on the sites being scraped, especially when nonstop automated crawlers hammer the servers.
2. Publishers argue that AI companies are now commercializing products built on datasets they never paid for, even as newsrooms themselves face declining revenue and heavy layoffs.

That tension is now visible everywhere, and the latest flashpoint is Wikipedia.

Wikipedia’s Costs Are Rising — and It Wants Compensation

Wikipedia co-founder Jimmy Wales said at the Reuters NEXT summit that the online encyclopedia is negotiating more deals with Big Tech to recover the financial burden created by AI companies training on its open-licensed content.

“The AI bots that are crawling Wikipedia are going across the entirety of the site,” Wales said. “So we have to have more servers, we have to have more RAM and memory for caching that, and that costs us a disproportionate amount.”

While Wikipedia’s text remains free for individual use — as required by its license — Wales drew a distinction between volunteers donating to keep the site running and multibillion-dollar corporations using Wikipedia as a backbone for commercial AI systems.

“Those people are donating money to support Wikipedia, and not to subsidize OpenAI costing us a ton of money. That doesn’t feel fair,” he said.

Wikipedia already signed a paid training-data deal with Google in 2022, and Wales confirmed that talks with other firms are ongoing. He didn’t rule out legal action against companies that continue training on Wikipedia without paying, saying that “soft power” shaming could be effective, but technical measures could also be deployed.

He mentioned Cloudflare’s AI Crawl Control, which allows websites to restrict how often and how deeply AI bots scrape their content. But that raises an ideological dilemma: limiting access contradicts Wikipedia’s long-standing commitment to free knowledge.

Still, Wales bluntly stated that the financial burden needs to be addressed.

Content Owners Strike Back

Between Meta’s new paid-news pipeline and Wikipedia’s push for licensing revenue, a new pattern is emerging in the AI ecosystem:
The era of “free training data” is ending.

Publishers — bruised by years of platform dominance over traffic and monetization — now see leverage in the AI boom. Professional newsrooms offer what models desperately need but cannot simulate:

• real-time reporting
• legally vetted information
• high-quality text at scale

That scarcity gives publishers negotiating power they didn’t have during the social-media era, when platforms controlled distribution and advertisers controlled revenue.

Nonprofits like the Wikimedia Foundation are now making the same argument: professional, volunteer-maintained knowledge bases are not cost-free inputs for trillion-dollar AI firms.

Why Meta’s Move Matters for the Industry

Meta’s new agreements signal three realities in the AI landscape:

1. AI chatbots cannot stay competitive without reliable, real-time information.
   Users expect AI systems to answer breaking-news questions with human-grade accuracy. Without licensing deals, models remain weeks or months behind real events.

2. Training-data scarcity is becoming an existential threat.
   As more websites block AI crawlers or demand payment, AI companies face a limited supply of high-quality material — especially news, science, medical content, and reference works.

3. Big Tech is quietly moving toward a “paid knowledge economy.”
   The model is shifting from scraping everything for free to selective licensing, with newsrooms and knowledge institutions charging for access.

This dynamic could reshape model training for years. The open-web era gave AI its early fuel; the next phase may look more like traditional media licensing — and more expensive.

What It Means Going Forward

Meta’s partnerships, arriving alongside Wikipedia’s hardening stance, highlight a deeper negotiation over who owns digital knowledge and how AI companies should pay for it. The industry’s biggest models have already devoured most freely available high-quality text online. From here on, AI companies seem to have two choices:

• pay for professionally published material
• or risk quality decline in their models

Meta’s decision suggests that Big Tech knows which way the wind is blowing.

And as AI adoption accelerates, the pressure will only grow. Publishers want a cut of the value their work generates for AI firms. Nonprofits want compensation for the cost of being mined. Regulators are asking who profits from public knowledge.