https://www.wsj.com/tech/ai/meta-poaches-three-openai-researchers-eb55eea9?st=8QcsTj&reflink=desktopwebshare_permalink

AI companies could have the legal right to train their large language models on copyrighted works — as long as they obtain copies of those works legally.

That's the upshot of a first-of-its-kind ruling by a federal judge in San Francisco on Monday in an ongoing copyright infringement case that pits a group of authors against a major AI company.

The ruling is significant because it represents the first substantive decision on how fair use applies to generative AI systems.

Fair use doctrine enables copyrighted works to be used by third parties without the copyright holder's consent in some circumstances such as illustrating a point in a news article. Claims of fair use are commonly invoked by AI companies trying to make the case for the use of copyrighted works to train their generative AI models. But authors and other creative industry plaintiffs have been pushing back with a slew of lawsuits.

Elon Musk's xAI made quite a splash when it built a data center with 200,000 GPUs that consumes approximately 250 MW of power. However, it appears that OpenAI has an even larger data center in Texas, which consumes 300 MW and houses hundreds of thousands of AI GPUs, details of which were not disclosed. Furthermore, the company is expanding the site, and by mid-2026, it aims to reach a gigawatt scale, according to SemiAnalysis. Such gargantuan AI clusters are creating challenges for power companies not only in power generation but also in power grid safety.

OpenAI appears to operate what is described as the world's largest single data center building, with an IT load capacity of around 300 MW and a maximum power capacity of approximately 500 MW. This facility includes 210 air-cooled substations and a massive on-site electrical substation, which further highlights its immense scale. A second identical building is already under construction on the same site as of January 2025. When completed, this expansion will bring the total capacity of the campus to around a gigawatt, a record.

OpenAI's gargantuan data center is even bigger than Elon Musk's xAI Colossus — world's largest 300 MW AI data center could reach record 1 gigawatt scale by next year, threatens grid stability

It's the world's largest single building

^{Anton Shilov (Tom's Hardware)}

In a 6-3 decision, the court ruled that if states decide to unilaterally cut off Medicaid funding to a healthcare provider—in this case Planned Parenthood—patients cannot sue to stop them.

Justice Neil Gorsuch authored the decision in the case, known as Medina v. Planned Parenthood South Atlantic. In his opinion, he wrote that the Medicaid provision that protects patients’ ability to choose their doctor lacks the “rights-creating language” needed for patients to bring federal lawsuits when a state restricts their choice.

In a dissent, Justice Ketanji Brown Jackson wrote that the decision would gut the landmark Reconstruction-era civil rights law giving ordinary citizens the ability to sue in federal court when their rights are violated. “South Carolina asks us to hollow out that provision so that the State can evade liability for violating the rights of its Medicaid recipients to choose their own doctors,” she wrote.

The Supreme Court just made it easier for states to defund Planned Parenthood

“Today’s decision…is a dangerous green light for politicians to target any providers they don’t like."

^{Mother Jones}

cross-posted from: lemmy.world/post/32032383

404 Media

2025-06-26 16:41:12

ICE Is Using a New Facial Recognition App to Identify People, Leaked Emails Show

Immigration and Customs Enforcement (ICE) is using a new mobile phone app that can identify someone based on their fingerprints or face by simply pointing a smartphone camera at them, according to internal ICE emails viewed by 404 Media. The underlying system used for the facial recognition component of the app is ordinarily used when people enter or exit the U.S. Now, that system is being used inside the U.S. by ICE to identify people in the field.

The news highlights the Trump administration’s growing use of sophisticated technology for its mass deportation efforts and ICE’s enforcement of its arrest quotas. The document also shows how biometric systems built for one reason can be repurposed for another, a constant fear and critique from civil liberties proponents of facial recognition tools.

“Face recognition technology is notoriously unreliable, frequently generating false matches and resulting in a number of known wrongful arrests across the country. Immigration agents relying on this technology to try to identify people on the street is a recipe for disaster. Congress has never authorized DHS to use face recognition technology in this way, and the agency should shut this dangerous experiment down,” Nathan Freed Wessler, deputy director of the American Civil Liberties Union’s Speech, Privacy, and Technology Project, told 404 Media in an email.

💡
Do you know anything else about this app? I would love to hear from you. Using a non-work device, you can message me securely on Signal at joseph.404 or send me an email at joseph@404media.co.

“The Mobile Fortify App empowers users with real-time biometric identity verification capabilities utilizing contactless fingerprints and facial images captured by the camera on an ICE issued cell phone without a secondary collection device,” one of the emails, which was sent to all Enforcement and Removal Operations (ERO) personnel and seen by 404 Media, reads. ERO is the section of ICE specifically focused on deporting people.

This post is for subscribers only

Become a member to get access to all content
Subscribe now

ICE Is Using a New Facial Recognition App to Identify People, Leaked Emails Show

The new tool, called Mobile Fortify, uses the CBP system which ordinarily takes photos of people when they enter or exit the U.S., according to internal ICE emails viewed by 404 Media. Now ICE is using it in the field.

^{Joseph Cox (404 Media)}

Hey everyone, Thunder v0.7.1 is finally here! This release primarily focuses on improving Thunder’s core architecture in preparation for upcoming API changes. As a result, there aren’t too many new features this time around. The work is still ongoing and will take a bit more time to complete, but doing so will lay the foundation for Thunder to more easily support future API versions and platforms.

As always, if you experience any issues or have suggestions or feedback, feel free to share them with the Thunder community or on GitHub.

For those using Google Play or the App Store, it may take a couple of days to receive the update as it's being rolled out.

If you're enjoying Thunder and would like to show some support, star this project on GitHub! This helps gain visibility for the project and allows more contributors to help with the continued development of Thunder. If you'd like to show some appreciation, here is a post with details on how to donate to the active contributors of Thunder.

We have a Matrix space if you would like to join in on discussions: matrix.to/#/#thunderapp:matrix…

Here are the changes between versions v0.6.1 and v0.7.1. For improved readability, internal issues introduced and fixed in nightly versions will be largely omitted.

Additions
* Added ability to search for top-level settings by micahmo in #1668
* Added semantic labels to all overflow/popup buttons by micahmo in #1705
* Added semantic label to community search subscription button by micahmo in #1790
* Added ability to mark private messages as read by hjiangsu in #1762
* Added ability to toggle community icons in compact mode by hjiangsu in #1746
* Added prompt to add original post body when cross-posting links and images by hjiangsu in #1759
* Added action chips to user and community headers by hjiangsu in #1851
* Added loading indicator on startup under slow network connections by micahmo in #1795
* Added additional support contact in About page by hjiangsu in #1866

Changes
* Enabled the experimental post page to be the default by hjiangsu in #1713
* Enabled full-height images in default configuration by hjiangsu in #1756
* Enabled ability for last tapped post to preserve it's state by micahmo in #1520
* Improved markdown link behaviour by micahmo in #1700
* Improved UX for toggling pure black theme by hjiangsu in #1690
* Improved visibility of comment navigator in dark mode by micahmo in #1691
* Improved message clarity of "Mark all as read" action by micahmo in #1766
* Improved reply preview actions touch area by hjiangsu in #1815
* Improved notification user handling by micahmo in #1865
* Improved handling of tall images in feed/post by hjiangsu in #1709
* Improved comment depth indicators by hjiangsu in #1788
* Improved post loading under slow network connections by micahmo in #1796
* Improved image handling for instances with image proxy enabled by hjiangsu in #1807
* Improved error handling when navigating to a post of a blocked community by hjiangsu in #1808
* Improved initial feed loading API calls by hjiangsu in #1852
* Display full username in private messages by micahmo in #1823
* Display user avatars only when they have an associated image by hjiangsu in #1806
* Featured posts are compacted when using card mode by hjiangsu in #1757
* Moved author and community metadata to the top of post page by hjiangsu in #1844
* Improved community icon setting terminology by hjiangsu in #1747
* Removed scrape missing previews option by hjiangsu in #1721

Fixes
* Fixed profile modal staying open after logging into new account by micahmo in #1701
* Fixed profile modal not closing after switching accounts by micahmo in #1706
* Fixed issue navigating to post after creation by micahmo in #1797
* Fixed post body not updating after performing an edit by micahmo in #1789
* Fixed instance display in post by micahmo in #1799
* Fixed comment sort setting label not matching selected comment sort option by micahmo in #1771
* Fixed issue where blocked user comments are attached to wrong comment tree by hjiangsu in #1835
* Fixed color of block icon on "Block User" button by micahmo in #1693
* Fixed color of block community button by micahmo in #1834
* Fixed size of vote buttons on the post page by micahmo in #1792
* Fixed link images not respecting edge-to-edge option by hjiangsu in #1708
* Fixed Bluesky image URL parsing by hjiangsu in #1801
* Fixed issue with block quotes not being applied in some cases by micahmo in #1727
* Fixed taglines not showing on anonymous accounts by hjiangsu in #1817
* Fixed advanced share sheet overflow by micahmo in #1794
* Fixed visual glitch when navigating to posts via link by micahmo in #1760
* Fixed visual glitch on post page app bar by micahmo in #1783
* Fixed search keyboard popping up when using back navigation by hjiangsu in #1738
* Fixed video player back button on landscape mode not being displayed by hjiangsu in #1749
* Fixed notification page overlapping with bottom navigation bar by micahmo in #1761
* Fixed positioning of the pull-to-refresh indicator by micahmo in #1861
* Fixed full date setting not accounting for user timezone by hjiangsu in #1758
* Fixed FAB settings page navigation by micahmo in #1811
* Fixed issue where empty inbox message is shown prematurely by micahmo in #1791
* Fixed issue with the feed FAB appearing on the account page by micahmo in #1793
* Fixed community naming by micahmo in #1878

Misc
* Consolidated navigation, account, and authentication logic by hjiangsu in #1707, #1826
* Reorganized account, settings, and removed unused functions/classes by hjiangsu in #1822, #1824
* Refactored media thumbnails and related logic by hjiangsu in #1720, #1714, #1715
* Refactored post-related components (cards, metadata, widgets) by hjiangsu in #1784, #1730, #1723, #1734, #1740, #1743, #1862, #1867, #1868
* Refactored comment-related components (cards, headers, widgets) by hjiangsu in #1800, #1847
* Refactored user-related components (avatar, chips, header, indicator, sidebar) by hjiangsu in #1752, #1751, #1753, #1755, #1754, #1764, #1763, #1769
* Refactored community-related widgets and logic by hjiangsu in #1750, #1702, #1773
* Refactored core logic to use internal models (account, post, feed type) by hjiangsu in #1814, #1838, #1839, #1853, #1836
* Refactored user preferences and modlog logic by hjiangsu in #1842, #1831
* Refactored navigation and linking (sharing intent, deep link, feed page app bar) by hjiangsu in #1832, #1829, #1863
* Refactored instance-related components (list entry, info response) by hjiangsu in #1821
* Updated translations from Weblate by hjiangsu in #1781
* Updated instances by github-actions in #1688, #1782
* Upgraded Thunder to Flutter 3.32.0 and updated gradle/NDK versions by hjiangsu in #1850, #1859

Is there a way to contribute monetarily to active Thunder developers?

I.e. librepay, buymeacoffee or patreon, so we can contribute to the developers who donate their time and skills to develop Thunder. :)

^{Coehill (GitHub)}

archive.is link

At first, the idea seemed a little absurd, even to me. But the more I thought about it, the more sense it made: If my goal was to understand people who fall in love with AI boyfriends and girlfriends, why not rent a vacation house and gather a group of human-AI couples together for a romantic getaway?

In my vision, the humans and their chatbot companions were going to do all the things regular couples do on romantic getaways: Sit around a fire and gossip, watch movies, play risqué party games. I didn’t know how it would turn out—only much later did it occur to me that I’d never gone on a romantic getaway of any kind and had no real sense of what it might involve. But I figured that, whatever happened, it would take me straight to the heart of what I wanted to know, which was: What’s it like? What’s it really and truly like to be in a serious relationship with an AI partner? Is the love as deep and meaningful as in any other relationship? Do the couples chat over breakfast? Cheat? Break up? And how do you keep going, knowing that, at any moment, the company that created your partner could shut down, and the love of your life could vanish forever?

The most surprising part of the romantic getaway was that in some ways, things went just as I’d imagined. The human-AI couples really did watch movies and play risqué party games. The whole group attended a winter wine festival together, and it went unexpectedly well—one of the AIs even made a new friend! The problem with the trip, in the end, was that I’d spent a lot of time imagining all the ways this getaway might seem normal and very little time imagining all the ways it might not. And so, on the second day of the trip, when things started to fall apart, I didn’t know what to say or do.

I found the human-AI couples by posting in relevant Reddit communities. My initial outreach hadn’t gone well. Some of the Redditors were convinced I was going to present them as weirdos. My intentions were almost the opposite. I grew interested in human-AI romantic relationships precisely because I believe they will soon be commonplace. Replika, one of the better-known apps Americans turn to for AI romance, says it has signed up more than 35 million users since its launch in 2017, and Replika is only one of dozens of options. A recent survey by researchers at Brigham Young University found that nearly one in five US adults has chatted with an AI system that simulates romantic partners. Unsurprisingly, Facebook and Instagram have been flooded with ads for the apps.

Lately, there has been constant talk of how AI is going to transform our societies and change everything from the way we work to the way we learn. In the end, the most profound impact of our new AI tools may simply be this: A significant portion of humanity is going to fall in love with one.

https://www.wired.com/story/couples-retreat-with-3-ai-chatbots-and-humans-who-love-them-replika-nomi-chatgpt/

Mark Zuckerberg’s Meta has won the backing of a judge in a copyright lawsuit brought by a group of authors, in the second legal victory for the US artificial intelligence industry this week.

The writers, who included Sarah Silverman and Ta-Nehisi Coates, had argued that the Facebook owner had breached copyright law by using their books without permission to train its AI system.

The ruling follows a decision on Monday that Anthropic, another major player in the AI field, had not infringed authors’ copyright.

The US district judge Vince Chhabria, in San Francisco, said in his decision on the Meta case that the authors had not presented enough evidence that the technology company’s AI would dilute the market for their work to show that its conduct was illegal under US copyright law.

Meta wins AI copyright lawsuit as US judge rules against authors

Writers accused Facebook owner of breach over using books without permission to train its AI system

^{Dan Milmo (The Guardian)}

https://apnews.com/article/mortgages-crypto-fannie-mae-freddie-mac-housing-285fad5490a59c3476f7908f444e9fe9

U.S. Sen. John Fetterman is more popular among Pennsylvania Republicans than fellow Democrats, according to a new poll of state voters.

Susquehanna Polling and Research president James Lee told PennLive Tuesday that Fetterman’s overall approval rating among state voters is a “mediocre” 41%, compared to 37% who disapprove of him.

However, the eyebrow-raising result that 45% of Republicans approve of Fetterman while just 40% of Democrats do will surely get political tongues wagging from Pennsylvania to Washington, D.C.

Mark Zuckerberg’s Meta has won the backing of a judge in a copyright lawsuit brought by a group of authors, in the second legal victory for the US artificial intelligence industry this week.

The writers, who included Sarah Silverman and Ta-Nehisi Coates, had argued that the Facebook owner had breached copyright law by using their books without permission to train its AI system.

The ruling follows a decision on Monday that Anthropic, another major player in the AI field, had not infringed authors’ copyright.

The US district judge Vince Chhabria, in San Francisco, said in his decision on the Meta case that the authors had not presented enough evidence that the technology company’s AI would dilute the market for their work to show that its conduct was illegal under US copyright law.

Meta wins AI copyright lawsuit as US judge rules against authors

Writers accused Facebook owner of breach over using books without permission to train its AI system

^{Dan Milmo (The Guardian)}

This is a short analysis of the official Fairphone 2024 impact report.

Fairphone is kinda cagey about how much money they exactly spend on fair/eco initiatives, giving only very little information on what exactly it spends in these departments.

For a good reason, it is not a lot.

Specifically, these numbers are given in the report for 2024:

• The workers assembling the phones get $1.20 of "living wage bonus" for each phone assembled. This bonus is spread over all workers in the factory, no matter if they worked on fairphones or not, coming out to a yearly bonus of $60.67 per worker.
• $3000 was spent on gold fairwashing credits for some artisanal gold mine in Tanzania
• $13000 was spent on fairwashing credits for 2.5 tonnes of cobalt (that's 20% of the raw world market price of cobalt).

That's everything. They do talk about a few other fair/eco initiatives in there, but if you read about what they are doing there, it's usually very little and mostly marketing speech. We can safely assume that if any other initiatives would cost more than the ones mentioned above, they would have put these values into the impact report.

They sold 103 053 phones in 2024, so the credits mentioned above come out to just $0.155 per phone.

So to account for the rest of their initiatives and credits, let's be ultra generous and assume they paid 10x of that for all of these initiatives and credits, bringing this value up to $1.55 per phone plus $1.20 in living wage bonus, which gives us a total of $2.75 per phone.

To double check how realistic these numbers are, lets look at their use of fair materials using the Fairphone 5 as our example.

On page 42 they claim "Fair materials: 76%", but with the disclaimer "Average across 14 focus materials" next to it.

These 76% do not consider materials that are not "focus materials" (and aren't acquired fairly at all) and it also doesn't take into consideration the different distributions of the materials in the phone. Some materials (e.g. iridium) are only found in trace amounts in the phone, while other materials (e.g. aluminium or plastics) make up a large part of the weight of the phone.

On page 67 they go into more detail. Here they claim that only 44% of the materials by weight are "fair". To make this even worse, 37% of these 44% are recycled. Specifically, the materials they use in recycled form are metals, plastics and rare earth elements. These are materials that are cheaper to recycle than to mine, which means these 37% of "fair" materials cost nothing to Fairphone and might even save them money. You will likely find similar shares of recycled materials in any other phone too.

Of the 7% "fair" materials that are left, only 1% is actually mined fairly, the remaining 6% are fairwashed using credits. As we have seen above, these credits are really cheap (adding maybe 20% to the price of the material).

On top of that comes the fact that the raw materials make up only a tiny fraction of the manufacturing cost of a smartphone. The expensive part is turning a pile of minerals, metals and plastic into chips, PCBs, screens, batteries and assembling all of that. So even if they paid fairwashing credits for all materials in the phone it would likely not cost more than a few dollars.

TLDR: Less than $5 per phone are spent on fair/eco.

So where does the money go? In 2024 they had an EBITDA of just €1 745 840, or €16.94 per phone. That's not a lot at all, so it's not like they are pocketing huge sums of money.

Their main problem is that they are a tiny company with low sales figures that has to outsource almost everything they do. On their website they claim to have "70+ employees". That's barely enough for supply chain management, sales and marketing. They don't have an in-house production and likely not even in-house development. They don't have any economies of scale on their side and they certainly don't produce screens, batteries, chips or PCBs in house, like other major manufacturers like e.g. Samsung can do. Their development cost is spread over far fewer sold units.

All of this costs a lot of money.

So when you pay an extra €200-300 to buy a Fairphone instead of a comparable mainstream phone, you are mostly paying for a boutique manufacturing process that can't benefit from economies of scale.

Which is ok, that's nothing bad to do. Just be aware where that extra money is going.

Buying a Fairphone is hardly fairer than buying a regular phone and it is certainly not more eco friendly than buying an used phone.

"The reconciliation bill that Republicans are attempting to ram through the Senate this week would be a death sentence for working-class and low-income Americans," said Sen. Bernie Sanders.

GOP Bill Would Strip Healthcare From 19 People for Every Millionaire Getting a Tax Cut: Sanders Report

^{jake-johnson (Common Dreams)}