On Friday people spotted and started sharing the Instagram profile for “Liv”, a “proud Black queer momma of 2 & truth-teller” and—with no apparent sense of irony—“your realest source for life’s ups & downs.” (If that’s true for you, you have my deepest sympathies.)

The Washington Post satirist Alexandra Petri summed things up quite well: “Everything about this new Meta AI bot is so dystopian but for me the AI-generated images of nonexistent donated coats is up there.” Also, there was the racist stereotyping in Liv’s chat replies.

The resulting outrage prompted Meta to quickly delete the “Liv” and a couple other profiles. However, the accounts were actually launched back in September 2023, and not quietly either. I don’t know what’s worse for Meta: the fact that the public still barely noticed these profiles until now, or the way in which Friday’s pushback has the potential to undermine Meta’s imminent AI plans. (Though if you know something I don’t about why fake profiles are actually a life-enhancing, brilliant idea, please do let me know.) —David Meyer

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

AI copyright settlement

The AI industry is afflicted with multiple copyright lawsuits over the provenance of its training data, but at least one has been partly resolved now.

Anthropic has settled some of the claims brought against it by music publishers (including the biggest, Universal) who said its Claude AI model was coughing up real lyrics from their catalogs, like chunks of “Uptown Funk”.

The publishers accused Anthropic of having “intentionally removed or altered copyright management information” from the data as they fed it into Claude as training material.

Now, Anthropic has agreed to, er, do what it’s already doing: applying training guardrails that are supposed to avoid copyright infringement. There is one potentially significant change, though—a new procedure that publishers can use to notify Anthropic if they see the guardrails aren’t working.

It remains to be seen whether the Tennessee federal court will force Anthropic to stop training future models on the lyrics. —DM

IBM and GlobalFoundries settle

Because a new year means a fresh start, IBM and GlobalFoundries have also settled their long-running legal battle over patents, trade secrets, and breach of contract.

IBM had sued GlobalFoundries on the basis that the spun-out former AMD chipmaking unit had messed up its server roadmap with production delays. GlobalFoundries then sued IBM for allegedly passing on its secret sauce to partners such as Intel and Rapidus.

That’s all now resolved…though we don’t know how, because the terms of the settlement are confidential. We do know that the two companies are now keen to “explore new opportunities for collaboration in areas of mutual interest.”

As The Register notes, the settlement should also cause a rare sigh of relief over at Intel HQ, as a GlobalFoundries win in that second suit could have caused further problems for the embattled chipmaker's roadmap.

Chinese phone fall

Things continue to worsen in China for Apple and other foreign smartphone companies, with their collective shipments to the country falling 47.4% year-over-year in November.

That’s worse than the 44.25% drop between October 2023 and October 2024. The data comes from China’s Academy of Information and Communications Technology (CAICT).

This is partly because people in China are switching to local brands, but also because the lackluster economy there is leading people to hold off on upgrades in general. As Reuters notes, shipments within China, local brands included, were down 5.1% year-over-year in November, dropping to 29.61 million devices.

Indeed, Beijing is so worried about the situation that it’s about to start subsidizing consumer purchases of phones, tablets and smartwatches. Whether foreign device makers like Apple will be included in the programs is unclear. 

India vs. VPNs

Around half a dozen VPN apps have reportedly disappeared from Apple and Google’s app stores in India, due to enforcement of a 2022 cybersecurity law.

The law states that VPN providers must store significant amounts of user data for five years—things like names and IP addresses and other stuff that rather undermines the privacy motivations of someone using a VPN.

But despite the fact that it’s a couple years old, the law hasn’t really been enforced until now. The apps that have been yanked from the Indian stores include heavy hitters like Cloudflare’s 1.1.1.1 and Hide.me.

According to the Times of India, rivals like NordVPN and ExpressVPN have loudly grumbled about the law and stopped marketing their services in India, but they still continue to serve customers there. Let’s see how long that lasts.

Russia vs. online maps

A Moscow court has reportedly ordered Yandex, Russia’s top search engine, to stop allowing people to look at online maps and images depicting Russian oil refineries.

The order, which follows a state request, comes in reaction to Ukraine’s use of drones to repeatedly attack one particular, unnamed refinery. Those attacks have caused severe damage and injury to refinery employees.

It should be noted that Russia’s military enthusiastically uses drones in Ukraine to hunt locals there, in what has grimly been referred to as “human safari.”

Yandex has long been known as “Russia’s Google”, although its holding company was for many years based in the Netherlands. Last July, that holding company (now called the Nebius Group) finally sold its Russian operations to a Russian consortium, which makes the service in that country more amenable than ever to the demands of the state.

More data

New evidence suggests Elon Musk is cosplaying his superfan Adrian Dittmann. Sad if true.

Tesla enters new year with laundry list of promises to fulfill after shocking market with its first-ever annual sales decline. Don’t hold your breath.

Computers of senior U.S. Treasury leaders accessed in hack. China calling.

Microsoft’s mini AI PCs are on the way. Time to compete with the new Mac Mini.

Apple auto-opts everyone into having their photos analyzed by AI for landmarks. Apple claims it’s respecting privacy.