Slashdot reader BrianFagioli writes: Florida International University researchers have developed a technique called JaiLIP (Jailbreaking with Loss-guided Image Perturbation) that uses subtle image modifications to bypass AI safety guardrails. Unlike traditional jailbreaks that rely on carefully crafted prompts, the attack works through images that appear normal to human viewers. The researchers tested the technique against BLIP-2, a multimodal AI model, and found that manipulated images significantly increased the likelihood of harmful responses. According to the study, the approach outperformed previous image-based jailbreak methods and nearly doubled the number of unsafe outputs generated during testing. The findings highlight a potential security risk for businesses deploying AI systems that process both images and text. While most discussions about AI safety focus on prompts, the research suggests that seemingly harmless images may also serve as an attack vector.

Read more of this story at Slashdot.

There's a deadly, record-breaking heat wave spreading east across Europe, reports the Washington Post — and it's even worse than a dire earlier forecast: The forecast was recorded in 2014 as part of a campaign coordinated by the World Meteorological Organization (WMO) that invited about 60 presenters worldwide to imagine a weather report from the year 2050. In one clip, Ãvelyne Dhéliat from French television network TF1 presented a hypothetical scenario of high temperatures 36 years into the future — during a heat wave in a warmer climate in 2050... One of the maps that Dhéliat shared was lit up in shades of orange, filled with temperature predictions of 40 degrees Celsius (104 degrees Fahrenheit), reaching as high as 43 degrees Celsius (109.4 degrees Fahrenheit). But it turns out, it didn't take 36 years for those imagined temperatures to be reached — and even exceeded. The heat on Wednesday alone, when the temperature soared as high as 112.3 degrees Fahrenheit (44.3 degrees Celsius), exceeded the 2050 projections in 19 out of 34 locations across mainland France — far sooner than some may have expected. Some places surpassed those hypothetical future temperatures by more than 20 degrees Fahrenheit. It's part of a dramatic shift in heat wave frequency across the country. Half of the heat waves observed since 1947 have occurred since 2010. "By 2100, heat waves could last up to two months continuously," the country's weather agency, Météo-France, said this week. It was hotter in France on Wednesday than in Las Vegas and Phoenix and just two degrees Fahrenheit shy of what was observed in Death Valley, California. An estimated less than one percent of the planet was hotter than France's hottest place... [T]he heat dome, which will linger into early next week, is only part of the story. This type of extreme heat is becoming more common as the planet warms, especially in Europe. Climate scientist Robert Rohde said in a post explaining the heat wave's causes that France and Western Europe should expect many more heat waves like this over the coming decades. "This isn't a fluke, but simply part of the new normal," he said. Thanks to Slashdot reader fjo3 for sharing the news.

Read more of this story at Slashdot.

Max Planck won 1918's Nobel Prize for physics. Yet two of his papers were retracted — a move now being criticized by Yves Gingras, a historian of physics at the University of Quebec and Mahdi Khelfaoui, a fellow historian of science at UQ Trois-Rivières. Science reports: The papers, both quietly retracted in 2011, originally appeared in the early 1940s in Naturwissenschaften, a German journal now owned by publishing giant Springer Nature. After some sleuthing, Khelfaoui determined one of the Planck pieces, a philosophical essay from 1942 titled "Sinn und Grenzen der exakten Wissenschaft" ("Meaning and Limits of Exact Science"), about how to achieve certainty in scientific knowledge, had also appeared in two other journals and been reprinted twice in books. Repackaging the same work multiple times is considered "self-plagiarism" and frowned upon today — the practice produces copyright conflicts and inflates scholars' publication records. The Naturwissenschaften site gives "copyright violation" as the reason for the retraction. Yet publishing identical material in multiple journals was widespread before the internet. "Science was more fragmented" then, Khelfaoui says. "You wanted different audiences ... to have access to your work." The practice was especially common for luminaries like Planck. Albert Einstein did the same (but escaped retractions). Springer Nature's "anachronistic" application of modern standards to a 1942 paper "distort[s] the historical record," Gingras and Khelfaoui argue in a preprint posted last month on arXiv. Any concerns about copyright violations are largely moot anyway: Because Planck died in 1947, his works are in the public domain in most countries. Gingras was especially incensed that Springer Nature deviated from the normal practice of merely slapping the word RETRACTED across the digital version of the paper while still allowing scholars to read the text. Instead, the publisher posted a blank white page with the cryptic phrase, "This article has been withdrawn due to article violation." Springer Nature is nevertheless still selling the empty PDF for $39.95. Suzanne Scarlata, a chemist and biochemist at the Worcester Polytechnic Institute and editor-in-chief of The Science of Nature, as Naturwissenschaften is now known, had not heard about the retractions before being contacted for this story... Scarlata suspects Springer Nature's internal policing software removed the paper and posted the retraction notice unilaterally, without human supervision: "I think it just happened with their algorithm," she says. "It's a mistake they should probably rectify." A second Planck paper was apparently removed because its response to a 1940 paper had used an identical title. Thanks to our long-time Slashdot reader He Who Has No Name for sharing the article.

Read more of this story at Slashdot.

When Mt. Vesuvius erupted in 79 A.D., it buried hundreds of papyrus scrolls. They were rediscovered in the mid-1700s, remembers Smithsonian magazine, "the only surviving collection of its kind from the Greco-Roman world..." "But when scholars tried to unroll them, the carbonized manuscripts crumbled to dust." Every generation that followed faced the same dilemma: They could wait for technology to advance, abandoning hope of reading the ancient texts in their own lifetime. Or they could try to open the scrolls themselves — and risk destroying them. In recent years, researchers have settled on a third option. Using advanced imaging and artificial intelligence, they're deciphering the scrolls without needing to unroll them at all. The Vesuvius Challenge has accelerated the process by turning it into a public competition, complete with cash prizes. In 2023, a student won $40,000 for deciphering a single word — "purple" — from an unopened scroll. Later, contestants would identify 2,000 Greek characters from one scroll ($700,000) and the title of another ($60,000). Now, for the very first time, researchers have recovered all surviving text from a single scroll. The nearly five-foot-long segment includes roughly 20 columns of ancient Greek philosophy, accessible for the first time in nearly 2,000 years. "The tech actually does look like magic, but it's not," Brent Seales, a computer scientist at the University of Kentucky, said at a press conference. (The article points out that Seales partnered with two Silicon Valley investors in 2023 to launch the Vesuvius Challenge, and is now hailing "the restoration of lost voices from the ancient world." Seales has been working on virtually unwrapping the scrolls since the early 2000s. The process involved imaging the bundles of papyrus using technology similar to CT scanners, isolating thin layers and then stitching them together.... "We've developed a systematic and a repeatable approach," Seales told the audience. "Now it's only a matter of time until we read all of the scrolls."

Read more of this story at Slashdot.

The Los Angeles Police Department says about 1,500 police agencies across America have drone programs, reports SFGate, and 58 of those drone-using police agencies are in California. The Sacramento County sheriff's office recently posted drone footage on Instagram set to theme from "Mission: Impossible," claiming "a nationwide first" where their drone successfully disarmed a felon "seen earlier with a firearm" (though now not moving, but holding a knife while lying face down in a garage). In the video the "not responding" suspect continues not moving as the drone dangles a magnet which catches on the knife. The drone then pulls multiple times until it comes out of the unmoving suspect's hand. The sheriff's office says their footage shows their drone "disarm an armed suspect, helping bring the incident to a safe resolution," in their post on Instagram, "rather than rush into a potentially deadly encounter..." Was he pretending to be dead or simply lying in wait for deputies to approach...? It's also worth noting that our drones are labeled as "military equipment" (even though anyone can purchase them at their local Walmart), but are really just another piece of technology helping deputies resolve dangerous situations safely. Their use protects both law enforcement personnel and suspects. SFGate offers more reports from around California: In Yucaipa, officials launched a Drone as First Responder (DFR) pilot program on May 28, the San Bernardino County Sheriff's Department announced this month. According to the release, drones have already been used to respond to over 100 calls for service, arriving before deputies for 71% of them. "The drones also contributed to 12 arrests, assisted in locating persons of interest on 37 occasions, and provided aerial overwatch during 44 incidents," it continues, though details on how they assisted the police are unclear. The drones, manufactured by Skydio, were also used to locate a young person experiencing a mental health crisis and another person launching illegal fireworks.

Read more of this story at Slashdot.

Linux kernel developer and Bytedance engineer Qi Zheng sent out a request for comments (RFC) patch series on a new feature called Reserved THP to combine the best of HugeTLB and THP kernel functionality...

The Jerusalem Post reports that doctors at Haifa's Rambam Health Care Campus "have successfully treated their first Israeli opioid addiction patient using an experimental noninvasive brain technology, easing him through withdrawal in just 20 minutes..." [T]he team of specialists at the Haifa medical center intervened in the electrical activity of an area of the patient's brain called the nucleus accumbens, the core of the brain system responsible for feelings of satisfaction, pleasure, and reward. The treatment, based on technology from the Israeli company Insightec, is similar to the one used to treat symptoms of essential tremor and Parkinsonian tremor, under MRI control. In this case, the treatment was carried out with the help of a new technology that performs noninvasive neuromodulation, without heating or burning tissue, and allows stimulation in the same area of the brain to increase or suppress activity... "Tests carried out a week later produced negative results for opioids and other substances," [said Dr. Lior Lev-Tov, director of the functional neurosurgery unit in Rambam's neurosurgery division and the one leading the new study at the medical center.] "The patient himself reported a craving score of zero out of 10 for using the drug, and even another side effect, a drastic drop in the desire for cigarettes, from three packs a day to just a few cigarettes, and with no urge to use alcohol. In other words, in a treatment that lasted about 20 minutes net, our patient was completely freed from an extreme dependence that had accompanied him every day for years. This is nothing less than a medical and therapeutic revolution." Dr. Lev-Tov added that "This experience opens doors for us to treat a wide range of very serious illnesses such as PTSD, OCD, eating disorders, other addictions, severe depression, severe pain disorders, and I hope we will also be able to reach cognitive areas and treat attention deficit disorders, Alzheimer's, Parkinson's, and more." Thanks to Slashdot reader Bruce66423 for sharing the article.

Read more of this story at Slashdot.

Thursday the Free Software Foundation blogged about this year's 47 'LibreLocal 2026' meetups, highlighting 10 that took place in Australia, Mexico, the United States, New Zealand, Cameroon, Switzerland, Spain, Argentina, China, and Iran. "Far from each other in many parts of the world, they came together around one unifying belief: free software." We envisioned LibreLocal as a collage of in-person community meetups that would bring people together to swap ideas, learn from each other, and celebrate free software. When we asked the free software community to organize LibreLocals last year, the response was very inspirational: 29 different meetups were hosted. After we made the global call this year, we were greeted with an even more enthusiastic response... Organizers hosted LibreLocals in cafes, bars, restaurants, libraries, universities, a computer repair shop, and even as part of a field trip to the System Source Museum, a museum dedicated to the history of computing in Hunt Valley, Maryland, USA. We also learned that a LibreLocal was organized inside Vakil Abad Prison in Mashhad, Iran by a free software supporter. Originally planned to be held in Shiraz, we were informed of this change in location on the LibreLocal wiki page set up for listing all LibreLocals. The updated entry, by another free software supporter in Iran, reads: "This year, one of our dedicated activists organized a LibrePlanet event from within prison in Iran. Currently serving a sentence for "cyber-crimes" related to his promotion of free software, he continues to introduce the principles of software freedom to his fellow inmates. We have placed this banner to honor his resilience and the community of individuals in prison who continue to stand for technological freedom. His identity will be revealed when it is safe to do so." Advocating for user freedom should never result in a prison sentence. We especially admire and respect the bravery and strength of those who fight for software freedom in the most dangerous and oppressive of environments. 50 people attended the LibreLocal meetup in Switzerland, according to one of the organizers, "forging connections between several local free software stakeholders and strengthening their cohesion." But the FSF's blog post stresses these are "ten stories among many more of free software supporters from across the globe... We also thank you our donors and associate members for the support that makes such meetups possible." The GNU Press Shop is now open through July 19 for their biannual fundraiser, offering a variety of freedom-respecting novelties including an FSF-branded antisurveillance webcam guard and both technical and philosophical books, like Richard Stallman's Free as in Freedom (which allegedly has turned up in Anthropic's training data). Other items include a slick new FSF logo sticker, a brass and zinc GNU "emblem" pin with real gold plating, and a cheeky sticker reminding everyone that "There is no cloud." And there's even a plush GNU toy.

Read more of this story at Slashdot.

An anonymous reader quotes a report from Business Insider: For the most powerful voices in AI, it's all about being in the loop. Claude Code creator Boris Cherny recently said he doesn't write his own AI prompts much anymore. Thanks to loops, he doesn't have to. "It's an agent that prompts Claude," Cherny recently told CNBC, adding, "I don't write the prompt anymore. Claude writes the prompt, and now I'm talking to that new Claude that is kind of coordinating." In the same interview, Cherny said that loops and a similar feature were examples of the kind of work he would be proudest of in a decade. Cherny isn't the only one embracing "loop engineering." OpenAI engineer Peter Steinberger, the creator of the viral OpenClaw project, wrote a public reminder to users who are still writing out prompts for AI agents. "Here's your monthly reminder that you shouldn't be prompting coding agents anymore," Steinberger wrote recently on X. "You should be designing loops that prompt your agents." [...] Steinberger shared an example of a loop he uses: "Tell codex to maintain your repos, wake up every 5 minutes and direct work to threads. That makes it easy to parallelize+steer work as needed." Claire Vo, founder of ChatPRD and host of the "How I AI," said, "it's really just reminding people that you don't have to use your human fingers to type in a prompt in order for your agent to do work on your behalf." The days of directly prompting generative AI coding tools are "kind of over, or at least some think it's going to be," Addy Osmani, director of Google Cloud, wrote in his post explaining the concept.

Read more of this story at Slashdot.

Wimbledon 2026 sur quelle chaîne en France ? C’est beIN Sports qui diffuse le tournoi en exclusivité du 29 juin au 12 juillet. Depuis votre décodeur b.tv, vous accédez à beIN Sports via l’offre Canal+ Sport ou en vous abonnant directement à beIN Sports depuis votre espace client Bouygues Telecom, pour suivre l’intégralité du tournoi […]

L'article Wimbledon 2026 sur quelle chaîne : Canal+ et Bbox a été publié en premier sur Bbox-Mag

Plage, piscine, sports nautiques : protéger son smartphone et ses accessoires de l’eau est devenu un réflexe indispensable pour les vacances. Housses étanches, supports waterproof, enceintes résistantes : notre sélection des meilleurs accessoires waterproof vacances disponibles en France pour profiter de l’été sans crainte. Pochettes et housses Supports vélo et moto Enceintes waterproof Écouteurs sport […]

L'article Accessoires waterproof vacances : coques et enceintes a été publié en premier sur Bbox-Mag

Que regarder ce week-end ? Voici notre sélection de films séries week-end pour le 27 et 28 juin. D’abord, la Finale du Top 14 en direct samedi. Ensuite, Scarface et Spectre dimanche. Enfin, deux sorties streaming majeures en toile de fond : The Bear saison 5 sur Disney+ et Avatar saison 2 sur Netflix. De […]

L'article Films séries week-end 27-28 juin : Top 14 et Scarface a été publié en premier sur Bbox-Mag

Un smartphone qui chauffe pendant la recharge, un jeu ou une vidéo en streaming, c’est fréquent. Une légère montée en température reste normale, mais une chaleur persistante ou excessive mérite attention. Elle peut signaler une application gourmande, une mauvaise habitude de charge ou, plus rarement, un problème matériel. Voici les causes les plus courantes et […]

L'article Smartphone qui chauffe : causes et solutions efficaces a été publié en premier sur Bbox-Mag

Regarder la finale Top 14 sur Bbox ne nécessite aucun équipement supplémentaire. Le samedi 27 juin au Stade de France, Canal+ diffuse ainsi la finale du TOP 14 2026 en exclusivité. Si vous êtes abonné Bbox, vous accédez directement à Canal+ depuis votre décodeur b.tv. Voici tout ce qu’il faut savoir pour profiter de la […]

L'article Finale Top 14 sur Bbox : Canal+ et décodeur b.tv a été publié en premier sur Bbox-Mag

Que voir sur CANAL semaine 22 juin ? La Finale du Top 14 en direct samedi au Stade de France, l’épisode 7 de Silo jeudi soir, et un catalogue riche avec Jurassic World Renaissance, Insaisissables 3, Dossier 137 ou encore Chien 51. La chaîne et la plateforme ont ainsi beaucoup à offrir cette semaine. Article […]

L'article CANAL semaine 22 juin : Finale Top 14 et Silo saison 2 a été publié en premier sur Bbox-Mag

Le Dolby Atmos est de plus en plus mis en avant sur les fiches techniques des barres de son, des téléviseurs et des services de streaming. Mais entre la promesse marketing et la réalité, la chaîne de conditions à réunir pour en profiter vraiment est souvent mal expliquée. Ce guide sur le Dolby Atmos streaming […]

L'article Dolby Atmos streaming : équipements, services et conditions réelles a été publié en premier sur Bbox-Mag

Pas de réseau, batterie qui chauffe, données mobiles épuisées : en vacances, votre GPS en ligne peut vous lâcher au pire moment. Les applis GPS hors ligne vacances fonctionnent sans connexion, avec des cartes téléchargées à l’avance, et certaines surpassent même Google Maps sur le terrain. Tour d’horizon des meilleures solutions disponibles en 2026. Pourquoi […]

L'article GPS hors ligne vacances : les meilleures applis 2026 a été publié en premier sur Bbox-Mag

Les nouveautés streaming semaine 22 arrivent en force, avec deux événements majeurs en tête : The Bear saison 5 sur Disney+ le 26 juin et Avatar saison 2 sur Netflix le 25 juin. La semaine démarre dès le dimanche 22 avec plusieurs lancements sur Prime Video et Disney+. Elle se poursuit ensuite chaque jour jusqu’au […]

L'article Nouveautés streaming semaine 22 : The Bear S5 et Avatar S2 a été publié en premier sur Bbox-Mag

Libéré, délivré

Nouvel épisode dans la saga Mythos. Anthropic a reçu la nuit dernière l’autorisation du gouvernement US pour déployer de nouveau Mythos 5 auprès d’organisations et entreprises états-uniennes. Fable 5 pourrait obtenir son feu vert la semaine prochaine.

Dévoilés le 10 juin, Fable 5 et Mythos 5, les modèles IA les plus avancés d’Anthropic, n’ont pas été très longtemps entre les mains des utilisateurs. Deux jours plus tard, Anthropic était obligée d’en fermer l’accès à tous les ressortissants étrangers ; étant dans l’impossibilité de déterminer la nationalité de ses utilisateurs, le labo IA a décidé d’interdire ces modèles à tout le monde.

• La Fable, le Mythos et la raison d’État

Depuis, les négociations se poursuivent entre l’administration Trump à l’origine de l’interdiction, et Anthropic. Elles ont fini par porter leurs fruits, du moins pour Mythos 5. Ce modèle est le frère jumeau de Fable 5, mais sans les restrictions d’usage liées à des sujets sensibles, comme la cybersécurité, la biologie, la chimie ou la distillation de modèles IA.

« Aujourd’hui, le gouvernement nous a informés que Mythos 5, notre modèle le plus avancé en cybersécurité, peut être redéployé auprès d’un ensemble d’organisations américaines chargées d’exploiter et de défendre des infrastructures critiques », a expliqué ce vendredi 26 juin l’entreprise sur les réseaux sociaux. L’accès est depuis en phase de rétablissement pour ces organisations.

Fable 5 dans l’antichambre

C’est un premier pas, mais insuffisant : Mythos 5, qui remplace Mythos Preview, se limite en effet aux membres du projet Glasswing. Fable 5 pourrait cependant suivre assez vite. Selon une indiscrétion d’Axios, l’administration Trump pourrait lever son interdiction sur le modèle grand public dans le courant de la semaine prochaine.

Howard Lutnick, le secrétaire au Commerce, a expliqué dans une lettre envoyée ce vendredi, et révélée par Semafor, qu’Anthropic avait travaillé avec le gouvernement sur les risques associés aux deux modèles. Des efforts qui ont permis « des avancées significatives ». L’entreprise s’est par ailleurs engagée à plancher sur « des protocoles, des normes et les modalités de mise à disposition des modèles ».

Cette « libération » des deux modèles sera vécue comme un soulagement par Anthropic et ses utilisateurs. Néanmoins, ce pataquès ne résout pas la demande du labo qui, le 12 juin, réclamait « un processus statutaire transparent, équitable, clair et fondé sur des faits techniques » plutôt que l’opacité actuelle qui « ne respecte pas ces principes ».

Un message similaire envoyé par OpenAI hier, alors que le créateur de ChatGPT lançait GPT-5.6… mais à une poignée restreinte d’organisations, comme l’a exigé le gouvernement. « Nous ne pensons pas que ce type de procédure d’accès gouvernementale doive devenir la norme à long terme », explique OpenAI. « Elle prive des meilleurs outils les utilisateurs, les développeurs, les entreprises, les défenseurs en cybersécurité et les partenaires internationaux qui en ont besoin. »

• OpenAI lance GPT-5.6 Sol, Terra et Luna, mais en accès limité

Libéré, délivré

Nouvel épisode dans la saga Mythos. Anthropic a reçu la nuit dernière l’autorisation du gouvernement US pour déployer de nouveau Mythos 5 auprès d’organisations et entreprises états-uniennes. Fable 5 pourrait obtenir son feu vert la semaine prochaine.

Dévoilés le 10 juin, Fable 5 et Mythos 5, les modèles IA les plus avancés d’Anthropic, n’ont pas été très longtemps entre les mains des utilisateurs. Deux jours plus tard, Anthropic était obligée d’en fermer l’accès à tous les ressortissants étrangers ; étant dans l’impossibilité de déterminer la nationalité de ses utilisateurs, le labo IA a décidé d’interdire ces modèles à tout le monde.

• La Fable, le Mythos et la raison d’État

Depuis, les négociations se poursuivent entre l’administration Trump à l’origine de l’interdiction, et Anthropic. Elles ont fini par porter leurs fruits, du moins pour Mythos 5. Ce modèle est le frère jumeau de Fable 5, mais sans les restrictions d’usage liées à des sujets sensibles, comme la cybersécurité, la biologie, la chimie ou la distillation de modèles IA.

« Aujourd’hui, le gouvernement nous a informés que Mythos 5, notre modèle le plus avancé en cybersécurité, peut être redéployé auprès d’un ensemble d’organisations américaines chargées d’exploiter et de défendre des infrastructures critiques », a expliqué ce vendredi 26 juin l’entreprise sur les réseaux sociaux. L’accès est depuis en phase de rétablissement pour ces organisations.

Fable 5 dans l’antichambre

C’est un premier pas, mais insuffisant : Mythos 5, qui remplace Mythos Preview, se limite en effet aux membres du projet Glasswing. Fable 5 pourrait cependant suivre assez vite. Selon une indiscrétion d’Axios, l’administration Trump pourrait lever son interdiction sur le modèle grand public dans le courant de la semaine prochaine.

Howard Lutnick, le secrétaire au Commerce, a expliqué dans une lettre envoyée ce vendredi, et révélée par Semafor, qu’Anthropic avait travaillé avec le gouvernement sur les risques associés aux deux modèles. Des efforts qui ont permis « des avancées significatives ». L’entreprise s’est par ailleurs engagée à plancher sur « des protocoles, des normes et les modalités de mise à disposition des modèles ».

Cette « libération » des deux modèles sera vécue comme un soulagement par Anthropic et ses utilisateurs. Néanmoins, ce pataquès ne résout pas la demande du labo qui, le 12 juin, réclamait « un processus statutaire transparent, équitable, clair et fondé sur des faits techniques » plutôt que l’opacité actuelle qui « ne respecte pas ces principes ».

Un message similaire envoyé par OpenAI hier, alors que le créateur de ChatGPT lançait GPT-5.6… mais à une poignée restreinte d’organisations, comme l’a exigé le gouvernement. « Nous ne pensons pas que ce type de procédure d’accès gouvernementale doive devenir la norme à long terme », explique OpenAI. « Elle prive des meilleurs outils les utilisateurs, les développeurs, les entreprises, les défenseurs en cybersécurité et les partenaires internationaux qui en ont besoin. »

• OpenAI lance GPT-5.6 Sol, Terra et Luna, mais en accès limité