Aggregator

AI seems to have a liberal bias

Amazon has asked the FCC to approve a 5,105-satellite constellation launching in 2028 that would provide global direct-to-cell "voice, messaging, data, and emergency services," similar to Starlink's partnership with T-Mobile. The network would use Globalstar spectrum, support future Apple satellite services, and complement Amazon's existing Leo broadband constellation. The Verge reports: Advanced Television points out that the FCC application also includes a request to use spectrum assigned to Iridium's satellite network outside the US, even as Rocket Lab is in the process of acquiring Iridium with plans to expand. The new constellation and direct-to-device service will "complement" the broadband services Amazon also plans to offer through its Leo platform. In addition to consumer satellite service, it would also be used for things like disaster response and emergency messaging. Amazon says the new direct-to-device satellites "will process the signals in orbit before relaying them to improve performance." The new constellation will be nearly double the size of Amazon's initial Leo constellation, which is planned to include around 3,200 satellites. As of July 2nd, Amazon says "over 375" of those satellites have launched so far.

Read more of this story at Slashdot.

American AI flag bearers get in line, at least on paper

Just don’t call the WinRT projection for Node.js a lock-in

wiredmikey shares a report from SecurityWeek: Nvidia and a large group of technology, cybersecurity, and enterprise software companies have launched new initiative aimed at developing and sharing open source tools, models, and techniques for securing AI systems and agents. The new Open Secure AI Alliance aims to give defenders more open tools for testing, auditing and protecting AI models and agents. Nvidia points to the recent security incident involving OpenAI and Hugging Face, noting that when closed AI tools could not differentiate between attackers and defenders and blocked forensic work, Hugging Face used the open-weight GLM 5.2 model on its own systems to review over 17,000 actions and contain the breach. "The right response is not to deny defenders access to capable open systems. It is to pair openness with strong safeguards, clear rules against malicious misuse, rigorous evaluation and rapid remediation. In cybersecurity, the safer path is the one that gives more defenders the ability to test, verify and strengthen the systems on which society relies," Nvidia said. Thanks to longtime Slashdot reader SphericalCrusher for also sharing the news.

Read more of this story at Slashdot.

Researchers find GLM and Kimi can adopt Claude's identity, but the evidence stops short of proving distillation

China too seems to be "on the move

Canonical introduces an optional virtualization HWE stack for Ubuntu 26.04 LTS, bringing newer QEMU, libvirt, EDK2, and SeaBIOS versions.

The new release also adds reproducible tarballs and new tools, while dropping Gold from the main tarball.

Longtime Slashdot reader AmiMoJo shares a report from Hackaday: Community-led open source project hosting site Codeberg has formally announced that projects whose code is largely or fully machine-generated through LLMs and other 'AI' tools will no longer be welcome. This follows on the heels of a similar ban on cryptocurrency-related projects. The community vote was on two issues, the first being the notion that scraping of project code for the use in LLMs should be forbidden, which was a motion that easily passed. The second motion was on disallowing projects whose code was substantially generated by LLMs like Claude, OpenAI Codex, and similar. This motion passed with 358 in favor versus 144 against. In the earlier linked blog post the reasoning behind especially this second issue is expanded upon, covering not only "license whitewashing," but also the direct and indirect hardware costs, with the expanding 'AI' datacenter hyperscaling having massively increased hardware costs for Codeberg over the past years, as the costs have been largely externalized. Also covered is the aspect of these LLM-based tools destroying the OSS community, which is something that is backed up by recent studies.

Read more of this story at Slashdot.

MDASH stuffed with MAI-Cyber-1-Flash and a side of GPT-5.4

According to The Information (paywalled), a state-backed Shanghai company has begun mass-producing China's first homegrown immersion DUV lithography machines, with about five units expected this year for SMIC, Hua Hong, and CXMT. Roughly 20 more units are expected in 2027. Tom's Hardware reports: The Information didn't name the manufacturer, but its sources described the operation as having pulled DUV development teams from several Chinese companies, one of them the state-backed startup Shanghai Yuliangsheng Technology. SMIC has been testing a Yuliangsheng immersion tool since September 2025. Most components in the new systems are domestic, though some critical parts still come from Japan, and delays at local suppliers have held back output this year. [...] China accounts for around 20% of ASML's net sales this year, down from 33% in 2025, driven mainly by mainstream logic demand... Independent analysis from the AI Futures Project in June put commercial-scale Chinese immersion DUV in the mid-2030s, with ASML holding 98.7% of the immersion market. Qualifying the new machines for production lines could take many months, and they trail ASML's tools on performance and build quality. China's domestic EUV effort, which Reuters first reported as a working prototype in December, remains years away.

Read more of this story at Slashdot.

An anonymous reader quotes a report from Ars Technica: OpenAI's ChatGPT is now refusing requests to generate text that directly mimics the style of famous authors. When asked to do so, the popular LLM instead offers a response that draws on the "broad qualities" of those authors "while remaining distinct in its own voice," for example. This morning, Ars received the following response to a test prompt asking for a story introduction in the style of Stephen King: "I can definitely write with the hallmarks of atmospheric, character-driven horror and small-town dread, but I can't write in Stephen King's exact style or closely imitate his distinctive voice. Here's an original opening that captures a similar feeling while remaining its own..." In testing, ChatGPT generated similar dodges for other authors both living (J.K. Rowling, Amy Tan) and dead (Charles Dickens, Ernest Hemingway). An analysis published by No Latency earlier this month (PDF) found the same behavior for living authors but found ChatGPT complied with style-copying requests for deceased authors. In refusing to directly copy the "exact style" of various authors, ChatGPT offered instead to capture an overall "feeling" by incorporating some of the common features found in those authors' work. That may seem like a distinction without a real difference at first glance. But the slight alteration could be legally important as OpenAI continues to fight a number of lawsuits brought by book authors alleging large-scale copyright infringement by models trained on their work. One of those suits specifically cites ChatGPT's "uncanny ability to generate text similar to that found in copyrighted textual materials," for instance. An OpenAI spokesperson did not respond to a request for comment from Ars Technica. In the US, copyright law generally protects only a specific expression of an idea, not the more intangible style of an author. But an AI-generated stylistic imitation could become infringing if it becomes "substantially similar" to the work of the original author. "We've never had a situation in which this personal style of individual creators could be imitated as well and as inexpensively as we now have with AI," George Washington University Law School Professor Robert Brauneis told Bloomberg Law.

Read more of this story at Slashdot.

MacRumors interviewed AI critic Ed Zitron, author of the Where's Your Ed At newsletter and host of the Better Offline podcast, about what could happen to Apple if the costly AI infrastructure boom collapses. Zitron argued that Apple is relatively well insulated because it has spent far less on data centers than rivals such as Microsoft, Google, and Amazon. He said the company could use the downturn to make selective acquisitions, though it might simply continue operating largely as before. "I think they will sit on the sidelines and watch everything burn," Zitron said. Here's an excerpt from the report: If the bubble deflates the way you expect (write-downs, too much compute supply, possibly OpenAI cratering), can you walk us through what that would actually look like for Apple? Does anything break for iPhone users, or does Apple mostly watch it happen from the sidelines? Could it even benefit Apple? I think things would look much the same for Apple. I think they will sit on the sidelines and watch everything burn. I could see them doing some choice acquisitions as things begin to collapse, but I could also see them do nothing. I think Apple is in a very weird place at the moment. The Vision Pro was a dud, but it was also the most interesting and future-forward thing I've seen anyone put out in a while. If they were smart, they'd tread water and sink as much money into making that as small as humanely possible -- no matter how long it takes -- because the entire AI bubble is a result of everybody running out of hypergrowth ideas, mostly because we're flat out of new interfaces. I want to be clear that what they want to do with the Vision Pro requires it to basically be weightless and invisible and never need adjustments. When it works, it's genuinely awesome. But that's a load-bearing when. One slight movement means the whole thing goes out of focus. I can't even use mine anymore because it needs an update that requires you to wear the thing the whole time. So much promise, released too early, shoved out the door by a CEO on his way out.

Read more of this story at Slashdot.

The Open Security AI Alliance says the Hugging Face/OpenAI mess proves frontier labs can't be trusted to properly secure sensitive systems

European misinformation researchers say TikTok, X, and Meta are obstructing access to platform data required under the EU's Digital Services Act through rejections, restrictive quotas, costly APIs, and burdensome security demands. Although regulators have fined X and pushed platforms to improve access, researchers remain skeptical that the changes will provide reliable, reproducible data at scale. Ars Technica reports: In recent years, social media companies shut down public access tools like Meta's CrowdTangle and replaced them with content libraries, or, like X, paywalled their API data, forcing academics to pay "hundreds of dollars a month," said Duncan Allen, a research officer at Democracy Reporting International (DRI) in Germany. Researchers say that without API access, what Iamnitchi describes as the "black holes" in what society knows about how these platforms recommend content or handle reports regarding sensitive content will only grow. Others, like TikTok, cap how many posts any researcher account can pull each day, which Allen said can make it "impossible to study anything at scale." The DSA was meant to solve this by allowing vetted researchers at credible institutions to have access to API data if they could show it would help in studying systemic risks, from illegal content to threats to fundamental rights. But two years after the law took effect, researchers say they struggle to meet its security requirements and face narrow interpretations from platforms of what qualifies as a systemic risk. Application forms differ by platform, but most require data to be stored on infrastructure that cannot be compromised -- such as a machine physically disconnected from the Internet -- a resource most universities lack, [said Adriana Iamnitchi, chair of computational social sciences at Maastricht University in the Netherlands, who leads research into online disinformation campaigns.] Even for researchers who secure approval, "there's no guarantee that the data is good," said L. K. Seiling, coordinator of the DSA40 Collaboratory, a German initiative that tracks 46 DSA applications. API data is often difficult for a colleague to reproduce, so a researcher's work cannot be checked for errors, which Iamnitchi said is a "basic requirement of science." DSA40 data shows that of 46 tracked applications, 20 were approved and 14 rejected. But approval rates vary widely: TikTok approved 11 of 13 applications, while X rejected 11 of 23. The true rejection rate is likely higher because the tracker relies on voluntary reporting, Seiling said. "There's no structured advantage for researchers to use this pathway," Seiling said. "Data access as it's set up right now tries to disincentivize researchers."

Read more of this story at Slashdot.

Catch up on the latest Linux news: COSMIC 1.4, VirtualBox 7.2.14, Firefox 153, Wine 11.14, Hyprland 0.56, OBS Studio 32.2, Debian developers debate ban on AI-assisted contributions, and more.

Is there a simulator for the wail when you realize your precious one-off bootleg has just been chewed up?

SPONSORED FEATURE: AI's hunger for data outstrips storage smarts, leaving GPUs famished

An anonymous reader quotes a report from Reuters: Nvidia is in talks to provide roughly $250 billion in financing guarantees for OpenAI as part of a massive data center project, the Wall Street Journal reported on Sunday. The backstop from Nvidia would help the ChatGPT maker lease a 10-gigawatt project that SoftBank's energy subsidiary is developing in southern Ohio, the newspaper said, citing people familiar with the matter. For OpenAI, a deal would be the first step toward controlling its own infrastructure instead of renting it from Microsoft, Amazon, and Oracle, while for Nvidia, it would guarantee demand for its chips for years to come. The project is expected to cost more than $500 billion in total, including the chips inside the data center, according to the WSJ. The $250 billion guarantee covers the data center lease and debt financing, but would not cover the Nvidia chips inside the center, the WSJ said, adding that the chipmaker was also discussing financing OpenAI's chip purchases worth up to $350 billion. Nvidia's backing would support financing vehicles aimed at reassuring lenders about the project's funding, the report added. The first phase of the project is expected to be finished in 2028, with around 800 megawatts of power, the Journal said. The report notes that the U.S. government will control access to the power, while Japan will fund it separately under a trade agreement tied to Tokyo's $33 billion investment in a natural gas plant. Commerce Secretary Howard Lutnick will reportedly help determine who receives access.

Read more of this story at Slashdot.