Apple's $50M bid to train AI on news

Training frontier AI requires massive amounts of data, and the tech giants are quickly realizing that quietly scraping the web is a legal minefield.

According to a report from The New York Times, Apple is actively negotiating with major news publishers to license their vast archives of articles, aiming to use that material to train its generative AI systems.

Apple is reportedly tabling multiyear deals worth at least $50 million. The company has already reached out to heavyweights including Condé Nast, NBC News, and media conglomerate IAC.

A $50M check meets mixed feelings

While big money is on the table, the publishing world's reception has reportedly been mixed.

Some executives are enthusiastic about forging a deep partnership with Apple. However, others are hesitant. The NYT notes that some publishers worry they could potentially be on the hook for legal liabilities resulting from Apple's use of their reporting.

To make matters more complicated, Apple has reportedly been vague about how exactly news content will be woven into its upcoming generative AI products. For now, Apple and the named publishers have declined to comment.

Apple is quietly burning millions a day

Apple has avoided the splashy consumer AI spotlight dominated by OpenAI, Microsoft, and Google, but it is moving aggressively behind closed doors.

Reports suggest the company is burning millions of dollars a day on AI research. It recently introduced a machine learning framework designed to run efficiently on Apple Silicon and has been heavily optimizing large language models to run natively on mobile hardware.

Bloomberg previously reported that Apple is overhauling Siri and prepping generative AI features for the next major iOS release. Apple isn't the first to strike licensing deals—OpenAI has already partnered with the Associated Press and Axel Springer—but Apple's scale makes this a notable move.

Why it matters

Apple is doing what Apple always does: moving quietly, waiting out the chaos, and paying upfront to avoid legal quagmires.

By offering $50 million multiyear checks, Apple is effectively building a clean, legally vetted data pipeline. Scrapers are getting sued across the industry, and we think Apple is deliberately spending its way out of future copyright litigation before its consumer features launch.

For publishers, this sets an expensive precedent. If Apple validates the idea that training data must be paid for at scale, rivals will face growing pressure to follow suit.

However, the publishers' skepticism is well-founded. Handing over decades of journalism to a notoriously secretive tech titan without knowing how it will be deployed is a gamble. Media companies risk feeding the very AI that could eventually replace their traffic.