News outfit AFP sues X/Twitter over the content it carries, and Elon Musk’s reaction is not wrong

Why does AFP get to claim licensing fees for this? Because of an increasingly popular concept known as neighboring rights or ancillary copyright, and if that sounds dull and obscure, it kind of is. But it’s also…well, here’s Musk’s take: “This is bizarre. They want us to pay *them* for traffic to their site where they make advertising revenue and we don’t!?”

Ancillary copyright was the brainchild of a cartel of German news publishers (led by Axel Springer, now the owner of Insider and Politico) that a decade ago managed to persuade the German government to pass a law allowing them to claim fees for the reproduction of their text snippets and image thumbnails by news aggregators, with Google being the main target. Spain then adopted its own “Google tax” and, long story short, Google managed to avoid paying anything in either country. So, to strengthen their hand, the publishers successfully lobbied politicians into making ancillary copyright an EU-wide thing in 2019. Google and Facebook now regularly pay publishers in countries such as France to carry bits of their content.

Australia followed suit in 2021 with its News Media Bargaining Code, but Google and Meta have so far managed to avoid being “designated” as platforms covered by that rulebook, by making a series of independent content-distribution deals with Australian news publishers that are reportedly worth more than $130 million a year. Canada also joined the party in June with its Online News Act, but that law’s relative inflexibility has led Meta to simply stop carrying news in Canada. Google has threatened to do the same.

I feel somewhat conflicted about ancillary copyright. On the one hand, the media is in crisis around the world and has been since the web became popular. Legacy publishers made money by carrying advertising, but the web opened them up to hordes of new competitors, while Google and then Facebook seized most of that ad value (which was partly the publishers’ fault, because they found it easier to hand over real estate on their web pages than to sell the spots themselves).

However—deep breath—Musk is right: It’s bizarre that those connecting news publishers with their readers have to pay the publishers to do so. There may be some kind of cosmic justice in seeing Google and Meta compensate those whose lunch they stole, but the mechanism itself is not logical—and the overall fairness of the transaction becomes far less clear when you’re talking about a platform like X/Twitter, which is not exactly a market leader in adtech.

Ancillary copyright is ultimately about Big Tech subsidizing the news media. Parts of the news media may need subsidizing—that’s a whole debate of its own—but if the platforms find the rules too onerous, they are within their rights to either stop carrying news (which has never been a big revenue driver for them) or to deprioritize it, as we have seen Facebook do over recent years. This is not a sustainable way to fund the news industry.

More of that pesky news stuff below.

Want to send thoughts or suggestions to Data Sheet? Drop a line here.

David Meyer

NEWSWORTHY

Russian hackers. Microsoft has accused a Russian hacker group called Midnight Blizzard (a.k.a. Cozy Bear or APT29) of being behind a run of “highly targeted social engineering attacks” that phished victims’ credentials in Teams chats. As the company noted in a blog post about the group’s methods, Midnight Blizzard is a Russian intelligence operation.

Alibaba follows Meta. Alibaba has open-sourced two large language models (LLMs). Releasing A.I.s into the wild like this is a first for a Chinese company, but it’s also what Meta just did with its LLaMa 2 model. The open-sourced Alibaba models are called Qwen-7B and Qwen-7B-Chat, and Reuters reports that they’re smaller versions of the company’s flagship Tongyi Qianwen LLM.

Prosecutors’ Binance conundrum. According to a Semafor report, the U.S. Justice Department is having trouble deciding whether or not to levy fraud charges against the Binance crypto exchange, because doing so might cause a run on the platform and spark general panic in crypto markets. As reporter Reed Albergotti points out: “The fact that the DOJ is debating the effects an indictment could have on consumers is, in a way, a nod to the legitimacy of crypto.”

SIGNIFICANT FIGURES

2

—The number of hours that Chinese under-18s would be allowed to spend on their smartphones each day, under draft rules laid out by the Cyberspace Administration of China. CNBC notes that this would have a big impact on local tech heroes like Tencent and ByteDance.

IN CASE YOU MISSED IT

Apple and Amazon, twin tech giants worth over $4 trillion, are about to bolster the economy’s upswing or spoil the party, by Stephen Pastis

The real reason Tesla beat its rivals in the charging wars, by Behnam Tabrizi

Furious Facebook Reels creators are discussing legal action after Meta reduced promised payouts by thousands of dollars due to a ‘display error,’ by Alexandra Sternlicht

Twitter CEO Linda Yaccarino changed her username to ‘LindayaX.’ Within moments, a troll account claimed ‘lindayaccs’ and took off, by Chris Morris

Billionaire Richard Branson’s Virgin Galactic saw a 400% jump in revenue, thanks to its space flights and ‘membership fees’ for astronauts-to-be, by Prarthana Prakash

BEFORE YOU GO

Meta’s music A.I. Meta has released a new A.I. Tool for turning text prompts into music or other audio. It’s called AudioCraft, and Meta insists that its text-to-music model, MusicGen, was trained on Meta-owned, specially licensed tracks, while the text-to-audio component, AudioGen, was trained on public sound effects. The package also includes an improved version of Meta’s EnCodec decoder, which sounds relatively free of the artifacts that make faux-music sound tinny and unreal.

The results sound passably authentic, in a background sort of way. Much as last year’s janky A.I. Imagery has given way to pictures that stand up to at least a few seconds’ scrutiny, the music and sounds that Meta’s models can now produce are far easier on the ears than what something like Dance Diffusion was generating in late 2022. It ain’t art, but it’s certainly technically impressive.