"For months, extremely personal and sensitive ChatGPT conversations have been leaking into an unexpected destination," reports Ars Technica: the search-traffic tool for webmasters , Google Search Console. Though it normally shows the short phrases or keywords typed into Google which led someone to their site, "starting this September, odd queries, sometimes more than 300 characters long, could also be found" in Google Search Console. And the chats "appeared to be from unwitting people prompting a chatbot to help solve relationship or business problems, who likely expected those conversations would remain private." Jason Packer, owner of analytics consulting firm Quantable, flagged the issue in a detailed blog post last month, telling Ars Technica he'd seen 200 odd queries — including "some pretty crazy ones." (Web optimization consultant Slobodan ManiÄ helped Packer investigate...) Packer points out "nobody clicked share" or were given an option to prevent their chats from being exposed. Packer suspected that these queries were connected to reporting from The Information in August that cited sources claiming OpenAI was scraping Google search results to power ChatGPT responses. Sources claimed that OpenAI was leaning on Google to answer prompts to ChatGPT seeking information about current events, like news or sports... "Did OpenAI go so fast that they didn't consider the privacy implications of this, or did they just not care?" Packer posited in his blog... Clearly some of those searches relied on Google, Packer's blog said, mistakenly sending to GSC "whatever" the user says in the prompt box... This means "that OpenAI is sharing any prompt that requires a Google Search with both Google and whoever is doing their scraping," Packer alleged. "And then also with whoever's site shows up in the search results! Yikes." To Packer, it appeared that "ALL ChatGPT prompts" that used Google Search risked being leaked during the past two months. OpenAI claimed only a small number of queries were leaked but declined to provide a more precise estimate. So, it remains unclear how many of the 700 million people who use ChatGPT each week had prompts routed to Google Search Console. "Perhaps most troubling to some users — whose identities are not linked in chats unless their prompts perhaps share identifying information — there does not seem to be any way to remove the leaked chats from Google Search Console.."

Read more of this story at Slashdot.

Le logiciel GPU-Tweak III et promet de prendre soin de la future ASUS ROG Matrix GeForce RTX 5090 Édition limitée 30e anniversaire, en surveillant son angle d'inclinaison, il faut dire qu'avec un tarif estimé à plus de 4000 euros, il vaut mieux en prendre soin ! En plus la carte s'annonce comme un joli pavé, avec ses dimensions de 370.3 x 150.5 x 77.3 mm et son occupation de 3.9 slots, on peut donc aisément imaginer qu'elle puisse entrainer une charge conséquente sur le PCB. […]

Lire la suite

On parle des prix des CPU et cette semaine cela bouge pas mal. On commence chez Intel avec le 14600K à son prix le plus bas, il perd 27 euros et s'affiche à 179 euros. Ensuite, le 14700K perd 9 euros, le 245K baisse de 7 euros, et enfin, le 265K de 5 euros. Chez AMD, le 7800X3D baisse fortement, il fait - 42 euros et s'affiche en plus au moins cher chez notre partenaire 1FODISCOUNT au prix de 315,90 euros, un excellent tarif. Ensuite, le 9700X baisse de 9 euros, le 9800X3D augmente de 26 euros et enfin, le 9950X3D perd 6 euros. […]

Lire la suite

Et c'est reparti pour une deuxième saison de montage et de watercooling en tout genre. Comme l'année dernière, nous ne parlons pas forcément de mode, mais surtout de beaux montages. Cette année, nous n'en ferons plus qu'un par semaine, le rythme fut parfois délicat à tenir. Comme toujours, n'hésitez pas à envoyer vos montages à Lucas. Ce dimanche, nous vous proposons de découvrir le Cougar Omnyx par GGF : […]

Lire la suite

The new series from Breaking Bad creator Vince Gilligan, Pluribus, was emphatically made by humans, not AI, reports TechCrunch: If you watched all the way to the end of the new Apple TV show "Pluribus," you may have noticed an unusual disclaimer in the credits: "This show was made by humans." That terse message — placed right below a note that "animal wranglers were on set to ensure animal safety" — could potentially provide a model for other filmmakers seeking to highlight that their work was made without the use of generative AI. In fact, yesterday the former X-Files writer told Variety "I hate AI. AI is the world's most expensive and energy-intensive plagiarism machine...." He goes on, about how AI-generated content is "like a cow chewing its cud — an endlessly regurgitated loop of nonsense," and how the U.S. will fail to regulate the technology because of an arms race with China. He works himself up until he's laughing again, proclaiming: "Thank you, Silicon Valley! Yet again, you've fucked up the world." He also says "there's a very high possibility that this is all a bunch of horseshit," according to the article. "It's basically a bunch of centibillionaires whose greatest life goal is to become the world's first trillionaires. I think they're selling a bag of vapor." And earlier this week he told Polygon that he hasn't used ChatGPT "because, as of yet, no one has held a shotgun to my head and made me do it." (Adding "I will never use it.") Time magazine called Thursday's two-episode premiere "bonkers." Though ironically, that premiere hit its own dystopian glitch. "After months of buildup and an omnipresent advertising campaign, Apple's much-anticipated new show Pluribus made its debut..." reports Macworld. "And the service promptly suffered a major outage across the U.S. and Canada." As reported by Bloomberg and others, users started to report that the service had crashed at around 10:30 p.m. ET, shortly after Apple made the first two episodes of the show available to stream. There were almost 13,000 reports on Downdetector before Apple acknowledged the problem on its System Status page. Reports say the outage was brief, lasting less than an hour... [T]here remains a Resolved Outage note on Apple TV (simply saying "Some users were affected; users experienced a problem with Apple TV" between 10:29 and 11.38 p.m.), as well as on Apple Music and Apple Arcade, which also went down at the same time. Social media reports indicated that the outage was widespread.

Read more of this story at Slashdot.

"The Firefox brand is getting a refresh and you get the first look," says a new web page at Firefox.com. "Kit's our new mascot and your new companion through an internet that's private, open and actually yours." Slashdot reader BrianFagioli believes the new mascot "is meant to communicate that message in a warmer, more relatable way." And Firefox is already selling shirts with Kit over the pocket (as well as stickers)...

Read more of this story at Slashdot.

For those dealing with SquashFS compressed, read-only file-systems, a new version of the user-space tools were released this week...

For more than a decade, the nonprofit Common Crawl "has been scraping billions of webpages to build a massive archive of the internet," notes the Atlantic, making it freely available for research. "In recent years, however, this archive has been put to a controversial purpose: AI companies including OpenAI, Google, Anthropic, Nvidia, Meta, and Amazon have used it to train large language models. "In the process, my reporting has found, Common Crawl has opened a back door for AI companies to train their models with paywalled articles from major news websites. And the foundation appears to be lying to publishers about this — as well as masking the actual contents of its archives..." Common Crawl's website states that it scrapes the internet for "freely available content" without "going behind any 'paywalls.'" Yet the organization has taken articles from major news websites that people normally have to pay for — allowing AI companies to train their LLMs on high-quality journalism for free. Meanwhile, Common Crawl's executive director, Rich Skrenta, has publicly made the case that AI models should be able to access anything on the internet. "The robots are people too," he told me, and should therefore be allowed to "read the books" for free. Multiple news publishers have requested that Common Crawl remove their articles to prevent exactly this use. Common Crawl says it complies with these requests. But my research shows that it does not. I've discovered that pages downloaded by Common Crawl have appeared in the training data of thousands of AI models. As Stefan Baack, a researcher formerly at Mozilla, has written, "Generative AI in its current form would probably not be possible without Common Crawl." In 2020, OpenAI used Common Crawl's archives to train GPT-3. OpenAI claimed that the program could generate "news articles which human evaluators have difficulty distinguishing from articles written by humans," and in 2022, an iteration on that model, GPT-3.5, became the basis for ChatGPT, kicking off the ongoing generative-AI boom. Many different AI companies are now using publishers' articles to train models that summarize and paraphrase the news, and are deploying those models in ways that steal readers from writers and publishers. Common Crawl maintains that it is doing nothing wrong. I spoke with Skrenta twice while reporting this story. During the second conversation, I asked him about the foundation archiving news articles even after publishers have asked it to stop. Skrenta told me that these publishers are making a mistake by excluding themselves from "Search 2.0" — referring to the generative-AI products now widely being used to find information online — and said that, anyway, it is the publishers that made their work available in the first place. "You shouldn't have put your content on the internet if you didn't want it to be on the internet," he said. Common Crawl doesn't log in to the websites it scrapes, but its scraper is immune to some of the paywall mechanisms used by news publishers. For example, on many news websites, you can briefly see the full text of any article before your web browser executes the paywall code that checks whether you're a subscriber and hides the content if you're not. Common Crawl's scraper never executes that code, so it gets the full articles. Thus, by my estimate, the foundation's archives contain millions of articles from news organizations around the world, including The Economist, the Los Angeles Times, The Wall Street Journal, The New York Times, The New Yorker, Harper's, and The Atlantic.... A search for nytimes.com in any crawl from 2013 through 2022 shows a "no captures" result, when in fact there are articles from NYTimes.com in most of these crawls. "In the past year, Common Crawl's CCBot has become the scraper most widely blocked by the top 1,000 websites," the article points out...

Read more of this story at Slashdot.

A CRISPR-based drug given to study participants by infusion is raising hopes for a much easier way to lower cholesterol, reports CNN: With a snip of a gene, doctors may one day permanently lower dangerously high cholesterol, possibly removing the need for medication, according to a new pilot study published Saturday in the New England Journal of Medicine. The study was extremely small — only 15 patients with severe disease — and was meant to test the safety of a new medication delivered by CRISPR-Cas9, a biological sort of scissor which cuts a targeted gene to modify or turn it on or off. Preliminary results, however, showed nearly a 50% reduction in low-density lipoprotein, or LDL, the "bad" cholesterol which plays a major role in heart disease — the No.1 killer of adults in the United States and worldwide. The study, which will be presented Saturday at the American Heart Association Scientific Sessions in New Orleans, also found an average 55% reduction in triglycerides, a different type of fat in the blood that is also linked to an increased risk of cardiovascular disease. "We hope this is a permanent solution, where younger people with severe disease can undergo a 'one and done' gene therapy and have reduced LDL and triglycerides for the rest of their lives," said senior study author Dr. Steven Nissen, chief academic officer of the Sydell and Arnold Miller Family Heart, Vascular & Thoracic Institute at Cleveland Clinic in Ohio.... Today, cardiologists want people with existing heart disease or those born with a predisposition for hard-to-control cholesterol to lower their LDL well below 100, which is the average in the US, said Dr. Pradeep Natarajan, director of preventive cardiology at Massachusetts General Hospital and associate professor of medicine at Harvard Medical School in Boston... People with a nonfunctioning ANGPTL3 gene — which Natarajan says applies to about 1 in 250 people in the US — have lifelong levels of low LDL cholesterol and triglycerides without any apparent negative consequences. They also have exceedingly low or no risk for cardiovascular disease. "It's a naturally occurring mutation that's protective against cardiovascular disease," said Nissen, who holds the Lewis and Patricia Dickey Chair in Cardiovascular Medicine at Cleveland Clinic. "And now that CRISPR is here, we have the ability to change other people's genes so they too can have this protection." "Phase 2 clinical trials will begin soon, quickly followed by Phase 3 trials, which are designed to show the effect of the drug on a larger population, Nissen said." And CNN quotes Nissen as saying "We hope to do all this by the end of next year. We're moving very fast because this is a huge unmet medical need — millions of people have these disorders and many of them are not on treatment or have stopped treatment for whatever reason."

Read more of this story at Slashdot.

A former Business Analyst reportedly filed a class action lawsuit claiming that for years, hundreds of remote employees at Bank of America first had to boot up complex computer systems before their paid work began, reports Human Resources Director magazine: Tava Martin, who worked both remotely and at the company's Jacksonville facility, says the financial institution required her and fellow hourly workers to log into multiple security systems, download spreadsheets, and connect to virtual private networks — all before the clock started ticking on their workday. The process wasn't quick. According to the filing in the United States District Court for the Western District of North Carolina, employees needed 15 to 30 minutes each morning just to get their systems running. When technical problems occurred, it took even longer... Workers turned on their computers, waited for Windows to load, grabbed their cell phones to request a security token for the company's VPN, waited for that token to arrive, logged into the network, opened required web applications with separate passwords, and downloaded the Excel files they needed for the day. Only then could they start taking calls from business customers about regulatory reporting requirements... The unpaid work didn't stop at startup. During unpaid lunch breaks, many systems would automatically disconnect or otherwise lose connection, forcing employees to repeat portions of the login process — approximately three to five minutes of uncompensated time on most days, sometimes longer when a complete reboot was required. After shifts ended, workers had to log out of all programs and shut down their computers securely, adding another two to three minutes. Thanks to Slashdot reader Joe_Dragon for sharing the article.

Read more of this story at Slashdot.

The Associated Press reports that "For the past decade, Dr. Priscilla Chan and her husband Mark Zuckerberg have focused part of their philanthropy on a lofty goal — 'to cure, prevent or manage all disease' — if not in their lifetime, then in their children's." During that decade they also funded other initiatives (including underprivileged schools and immigration reform), according to the article. But there's a change coming: Now, the billionaire couple is shifting the bulk of their philanthropic resources to Biohub, the pair's science organization, and focusing on using artificial intelligence to accelerate scientific discovery. The idea is to develop virtual, AI-based cell models to understand how they work in the human body, study inflammation and use AI to "harness the immune system" for disease detection, prevention and treatment. "I feel like the science work that we've done, the Biohub model in particular, has been the most impactful thing that we have done. So we want to really double down on that. Biohub is going to be the main focus of our philanthropy going forward," Zuckerberg said Wednesday evening at an event at the Biohub Imaging Institute in Redwood City, California.... Chan and Zuckerberg have pledged 99% of their lifetime wealth — from shares of Meta Platforms, where Zuckerberg is CEO — toward these efforts... On Thursday, Chan and Zuckerberg also announced that Biohub has hired the team at EvolutionaryScale, an AI research lab that has created large-scale AI systems for the life sciences... Biohub's ambition for the next years and decades is to create virtual cell systems that would not have been possible without recent advances in AI. Similar to how large language models learn from vast databases of digital books, online writings and other media, its researchers and scientists are working toward building virtual systems that serve as digital representations of human physiology on all levels, such as molecular, cellular or genome. As it is open source — free and publicly available — scientists can then conduct virtual experiments on a scale not possible in physical laboratories. "We will continue the model we've pioneered of bringing together scientists and engineers in our own state-of-the-art labs to build tools that advance the field," according to Thursday's blog post. "We'll then use those tools to generate new data sets for training new biological AI models to create virtual cells and immune systems and engineer our cells to detect and treat disease.... "We have also established the first large-scale GPU cluster for biological research, as well as the largest datasets around human cell types. This collection of resources does not exist anywhere else."

Read more of this story at Slashdot.

Merged today ahead of the Linux 6.18-rc5 kernel due out on Sunday is a partial fix for a performance regression observed on IBM POWER hardware...

C'est une vidéo qui sort un petit peu de nos sentiers battus que nous vous proposons ce soir. Nous sommes toujours dans du "H&Co", en déviant un peu vers le hardware des smartphones, mais nous pensons que le test dont nous allons parler pourrait tout de même intéresser beaucoup d'entre vous, car...

Long-time Slashdot reader AmiMoJo shared this report from Marine Insight: The world's largest cargo sailboat, Neoliner Origin, completed its first transatlantic voyage on 30 October despite damage to one of its sails during the journey. The 136-metre-long vessel had to rely partly on its auxiliary motor and its remaining sail after the aft sail was damaged in a storm shortly after departure... Neoline, the company behind the project, said the damage reduced the vessel's ability to perform fully on wind power... The Neoliner Origin is designed to reduce greenhouse gas emissions by 80 to 90 percent compared to conventional diesel-powered cargo ships. According to the United Nations Conference on Trade and Development (UNCTAD), global shipping produces about 3 percent of worldwide greenhouse gas emissions... The ship can carry up to 5,300 tonnes of cargo, including containers, vehicles, machinery, and specialised goods. It arrived in Baltimore carrying Renault vehicles, French liqueurs, machinery, and other products. The Neoliner Origin is scheduled to make monthly voyages between Europe and North America, maintaining a commercial cruising speed of around 11 knots.

Read more of this story at Slashdot.

"Internal documents have revealed that Meta has projected it earns billions from ignoring scam ads that its platforms then targeted to users most likely to click on them," writes Ars Technica, citing a lengthy report from Reuters. Reuters reports that Meta "for at least three years failed to identify and stop an avalanche of ads that exposed Facebook, Instagram and WhatsApp's billions of users to fraudulent e-commerce and investment schemes, illegal online casinos, and the sale of banned medical products..." On average, one December 2024 document notes, the company shows its platforms' users an estimated 15 billion "higher risk" scam advertisements — those that show clear signs of being fraudulent — every day. Meta earns about $7 billion in annualized revenue from this category of scam ads each year, another late 2024 document states. Much of the fraud came from marketers acting suspiciously enough to be flagged by Meta's internal warning systems. But the company only bans advertisers if its automated systems predict the marketers are at least 95% certain to be committing fraud, the documents show. If the company is less certain — but still believes the advertiser is a likely scammer — Meta charges higher ad rates as a penalty, according to the documents. The idea is to dissuade suspect advertisers from placing ads. The documents further note that users who click on scam ads are likely to see more of them because of Meta's ad-personalization system, which tries to deliver ads based on a user's interests... The documents indicate that Meta's own research suggests its products have become a pillar of the global fraud economy. A May 2025 presentation by its safety staff estimated that the company's platforms were involved in a third of all successful scams in the U.S. Meta also acknowledged in other internal documents that some of its main competitors were doing a better job at weeding out fraud on their platforms... The documents note that Meta plans to try to cut the share of Facebook and Instagram revenue derived from scam ads. In the meantime, Meta has internally acknowledged that regulatory fines for scam ads are certain, and anticipates penalties of up to $1 billion, according to one internal document. But those fines would be much smaller than Meta's revenue from scam ads, a separate document from November 2024 states. Every six months, Meta earns $3.5 billion from just the portion of scam ads that "present higher legal risk," the document says, such as those falsely claiming to represent a consumer brand or public figure or demonstrating other signs of deceit. That figure almost certainly exceeds "the cost of any regulatory settlement involving scam ads...." A planning document for the first half of 2023 notes that everyone who worked on the team handling advertiser concerns about brand-rights issues had been laid off. The company was also devoting resources so heavily to virtual reality and AI that safety staffers were ordered to restrict their use of Meta's computing resources. They were instructed merely to "keep the lights on...." Meta also was ignoring the vast majority of user reports of scams, a document from 2023 indicates. By that year, safety staffers estimated that Facebook and Instagram users each week were filing about 100,000 valid reports of fraudsters messaging them, the document says. But Meta ignored or incorrectly rejected 96% of them. Meta's safety staff resolved to do better. In the future, the company hoped to dismiss no more than 75% of valid scam reports, according to another 2023 document. A small advertiser would have to get flagged for promoting financial fraud at least eight times before Meta blocked it, a 2024 document states. Some bigger spenders — known as "High Value Accounts" — could accrue more than 500 strikes without Meta shutting them down, other documents say. Thanks to long-time Slashdot reader schwit1 for sharing the article.

Read more of this story at Slashdot.

Long-time Slashdot reader AmiMoJo shared this report from Linuxiac: The Japanese branch of Mozilla's Support Mozilla (SUMO) community — responsible for localizing and maintaining Japanese-language support documentation for Firefox and other Mozilla products (consisting of Japanese native speakers) — has officially disbanded after more than two decades of voluntary work... SUMO, short for Support Mozilla, is the umbrella project for Mozilla's user support platform, support.mozilla.org, that brings together volunteers and contributors worldwide who translate, maintain, and update documentation, tutorials, and troubleshooting guides for Firefox, Thunderbird, and other Mozilla products... According to marsf, the long-time locale leader of the Japanese SUMO team, the decision to disband was triggered by the recent introduction of an automated translation system known as Sumobot. Deployed on October 22, the bot began editing and approving Japanese Knowledge Base articles without community oversight. The article notes marsf's complaints in a post to the SUMO discussion forum, including the fact that the new automated system automatically approved machine-translated content with only a 72-hour window for human review. As a result, more than 300 Knowledge Base articles were overwritten on the production server, which marsf called "mass destruction of our work."

Read more of this story at Slashdot.

L’écran Titan Army P275MV PLUS propose une dalle de 27″ de diagonale en Mini-LED très lumineuse. La définition est de 3840 x 2160 pixels avec un rafraichissement de 160 Hz et peut atteindre 320 Hz en 1920 x 1080 pixels. La luminosité grimpe à 450 nits en SDR et 1000 nits en HDR. On retrouve une colorimétrie exemplaire à 153% de la norme sRGB et 96% du DCI-P3. Plusieurs scénarios d’affichage sont disponibles pour le jeu. Des fonctions PIP et PBP sont présentes pour piloter plusieurs sources à l’écran. Le pied offre des réglages ergonomiques complets. La connectique liste deux entrées HDMI 2.1, un DisplayPort 1.4 et un USB Type-C avec une alimentation 65 Watts. Un hub USB double port de type-A et un Jack audio sont présents ainsi qu’une paire d’enceintes intégrées.

L’écran est proposé à 379.99€ et baisse à 339.99€ avec le code 25NOVP1. La livraison du Titan Army P275MV PLUS sera gratuite depuis l’Allemagne.

Voir l’offre sur Geekbuying

Bon Flan : Ecran 27″ Titan Army P275MV PLUS Mini-LED à 359€ 🍮 © MiniMachines.net. 2025

Le record du monde d'overclocking DDR5 a une fois encore été officiellement battu en ce 8 novembre 2025. Il y a seulement un mois et demi de cela, le seuil des 13 000 MT/s était atteint pour la toute première fois de manière officieuse, car la validation CPU-Z n'avait pas été publiée par SaltyCroiss...

American multinational freight company UPS "has grounded its fleet of MD-11 aircraft," reports the Guardian, "days after a cargo plane crash that killed at least 13 people in Kentucky. The grounded MD-11s are the same type of plane involved in Tuesday's crash in Louisville. They were originally built by McDonnell Douglas until it was taken over by Boeing." More details from NBC News: UPS said the move to temporarily ground its MD-11 fleet was made "out of an abundance of caution and in the interest of safety." MD-11s make up 9% of the company's air fleet, it said. "We made this decision proactively at the recommendation of the aircraft manufacturer. Nothing is more important to us than the safety of our employees and the communities we serve," UPS spokesman Jim Mayer said... FedEx said early Saturday that it was also grounding its MD-11s. The UPS rival has 28 such planes in operation, out of a fleet of around 700, FedEx said. Video shows that the left engine of the plane caught fire during takeoff and immediately detached, National Transportation Safety Board member Todd Inman said Wednesday. The National Transportation Safety Board is the lead agency in the investigation. Thanks to long-time Slashdot reader echo123 for suggesting the article.

Read more of this story at Slashdot.

Ryzen AI Software as AMD's collection of tools and libraries for AI inferencing on AMD Ryzen AI class PCs has Linux support with its newest point release. Though this "early access" Linux support is restricted to registered AMD customers...