❌

Vue normale

Reçu — 22 septembre 2026 ⏭

Claude Opus 5.5 et GPT-6 Sol sortent le même jour : Anthropic et OpenAI ont du mal à « ralentir » l’IA

✇Numerama.com - Magazine
Par : Nicolas Lellouche
22 septembre 2026 à 20:11

Quelques jours après avoir appelé à ralentir la course à l'IA, Anthropic lance Claude Opus 5.5 et OpenAI complète sa gamme GPT-6 avec GPT-6-Sol et GPT-6-Luna. La bonne surprise vient des prix : les deux entreprises baissent les tarifs de leurs API.

2h de cours par jour grâce à l’IA : c’est quoi Alpha School, l’école choisie par Fidji Simo pour sa fille ?

✇Numerama.com - Magazine
Par : Lisa Imperatrice
22 septembre 2026 à 14:26

Fidji Simo, entrepreneuse française et ancienne patronne des applications d’OpenAI, vient d’inscrire sa fille à Alpha School. Ce réseau américain d’écoles privées promet de concentrer les matières fondamentales en seulement deux heures par jour, notamment grâce à l'IA.

Pentagon Investigators Say Overreliance on Palantir AI Contributed to US Strike That Killed 123 Iranian Children

✇Slashdot
Par : EditorDavid
22 septembre 2026 à 13:04
Two U.S. missiles hit an elementary school in Iran, killing over 150 people including 123 children on the first day of the Iran War. But Gizmodo notes that a new Bloomberg investigation cites U.S. officials involved in an unreleased internal Pentagon review who blame "a cascade of preventable failures that included an overreliance on an AI tool built by Palantir." According to the officials who spoke to Bloomberg, some personnel inside U.S. Central Command leaned too hard on the AI inside the Maven Smart System, an AI-powered data integration and targeting platform developed by Palantir to accelerate intelligence analysis and decision-making. The Defense Department has reportedly made the software tool a cornerstone of military operations over the past year... Officials said some users expected Maven to flag stale records or contradictions in the intelligence assembled for potential targets, though it is not clear why they thought the system would do that. The Minab [elementary school] site, which was cataloged as an Islamic Revolutionary Guard Corps facility due to outdated data, was fed into Maven with other candidates and came out as a recommended day-one target. Target-list work that once took hours was condensed into minutes. A Palantir spokesperson told Bloomberg the company "is not responsible for the underlying data nor identifying intelligence deficiencies" and that there is no evidence its software was at fault. Two people familiar with Palantir's Pentagon contracts said the government keeps primary responsibility for the quality of the information that's fed into Maven. After the strike, Palantir added features that "re-review underlying intelligence to identify factors that would disqualify a target and flag inconsistencies and inaccuracies that human review may have missed," a person familiar with the work said. That new functionality is said to have already caught some anomalies... More than 1,000 Iranian targets were hit in the first 24 hours, and according to Bloomberg's sources, that pace compressed the time available for additional confirmations. Staffing on civilian harm mitigation teams across the U.S. Department of Defense had also fallen by roughly 90% over recent years, shrinking to fewer than 20 people overall. The U.S. Central Command's group shrank from 10 people to one. No member of those teams reviewed the Minab site before the missiles were up in the air. A separate inquiry by the United Nations' Independent International Fact-Finding Mission on Iran this week reached the conclusion that there were reasonable grounds to believe the United States committed the war crime of launching an indiscriminate attack.

Read more of this story at Slashdot.

Face à la colère des médaillés Fields, OpenAI s’entoure de neuf grands chercheurs pour tenter d’apaiser les mathématiciens

✇Numerama.com - Magazine
Par : Julien Lausson
22 septembre 2026 à 09:20

maths openai

OpenAI a annoncé ce lundi 21 septembre 2026 la création d'un groupe consultatif de mathématiciens de renom, chargé de l'aider à publier ses résultats sans froisser la profession. Une main tendue après des semaines de tensions autour de ses annonces en maths.

Meta's New 'Personal AI Agent' Muse Beats ChatGPT in Downloads - and Get Blocked by Amazon

✇Slashdot
Par : EditorDavid
22 septembre 2026 à 04:04
Meta's AI "personal agent" Muse overtook ChatGPT as the #1 iOS free app in the U.S. on Friday, reports CNBC — and it's still showing more downloads as of Monday night. Muse now has over 2.5 million downloads since its debut less than two weeks ago, according to analytics firm Sensor Tower... [Muse] is currently ahead of popular services like OpenAI's ChatGPT, Polymarket and Kashi... It's also ahead of rival AI apps like Anthropic's Claude and SpaceXAI's Grok AI, in addition to the Facebook-parent's own Meta AI app... Meta pitched the app as a way for consumers to manage and direct supercharged digital assistants that can perform a number of tasks across the web, like filling out electronic forms and organizing email inboxes... "I think up to this point most of the agentic use cases have not really been for normal people," Bernstein analyst Stacy Rasgon told CNBC in a "Squawk on the Street" interview on Monday. "Now you've got Meta's Muse and other agents out there that are starting to maybe get more potential for more broad-based adoption." The buzz around Meta was also reflected in the stock, which surged more than 11% on Monday. Amazon is already blocking Muse, reports GeekWire. Amazon first tried to get Meta to voluntarily exclude Amazon, but they were unsuccessful: The problem, Amazon says, is that it never agreed to any of it. Meta didn't tell Amazon that Muse would access its store, the agent doesn't identify itself when it browses, and it appears to capture and store customer credentials, which the company says could create privacy and security risks. As of Sunday night, people trying to use Muse to shop on Amazon were seeing the popup, "Continued access by an unauthorized AI agent violates Amazon's Conditions of Use, to which our customers have agreed..." Amazon has spent the past year trying to keep outside agents off its site, suing Perplexity over its Comet browser and moving to block shopping agents from Google and OpenAI. [Amazon won a preliminary injunction against Perplexity in March, the article poitns out, "then lost it on Aug. 4, when the Ninth Circuit ruled that the user — not the AI company — was the one accessing Amazon's computers under federal anti-hacking law. The court denied Amazon's petition for rehearing on Sept. 10."] On Sunday night, Amazon said it is in direct conversation with Meta about the issue. Asked whether it would consider taking legal action, the company declined to comment. The business stakes are high for Amazon. In addition to operating its flagship e-commerce site, Amazon generated more than $68 billion in ad revenue last year — a business that depends on people browsing its pages and seeing sponsored products. Muse also appears to scrape account data, reports CNBC. But Shopify, however, "is working with Meta on Muse access. Shopify CEO Tobias Lütke announced Monday that the e-commerce site was partnering with Muse to allow agentic checkout in its online stores."

Read more of this story at Slashdot.

Reçu — 21 septembre 2026 ⏭

Essai Opel Corsa GSE : le retour de la sportive électrique

✇le blog auto
Par : David Christian Koskas
21 septembre 2026 à 16:09

Dix ans après l’arrêt de la Corsa OPC, Opel ressuscite sa citadine sportive, et le fait avec un badge inédit : GSE, pour Grand Sport Electric. Après le Mokka GSE l’an dernier, c’est au tour de la Corsa de recevoir le traitement haute performance version zéro émission. Présentée en mai dernier et dévoilée aux médias en août, elle fera sa première apparition publique au Mondial de l’Automobile de Paris en octobre, avec des commandes ouvertes dès septembre.

Une citadine qui joue des coudes

Difficile de confondre cette Corsa avec une version de base. Le bouclier avant spécifique GSE attire immédiatement le regard, tandis que l’arrière adopte un traitement tout aussi affirmé, avec des inserts noirs qui courent sur le bas de caisse et autour des passages de roues élargis. Le toit et le spoiler arrière passent au noir, en contraste marqué avec la carrosserie, et les jantes de 18 pouces à trois branches, chaussées de Michelin Pilot Sport 4S en 215/40, laissent entrevoir des étriers de frein Alcon peints en jaune et floqués du logo GSE. Une signature visuelle qui ne cache rien de ses ambitions.

Un habitacle qui rend hommage à la Corsa GSi

À bord, Opel a soigné les clins d’œil aux Corsa sportives des années 80 et 90. Les sièges Performance, garnis d’Alcantara et floqués du logo GSE, reprennent un revêtement à damier noir, gris et jaune directement inspiré des anciennes GSi, tandis que les ceintures de sécurité jaune vif viennent trancher avec le reste de l’habitacle. Le volant est lui aussi gainé d’Alcantara, les pédales passent à l’aluminium, et la planche de bord arbore fièrement son logo GSE sur fond gris argenté. Niveau technologie, l’écran tactile central de 10 pouces embarque un mode dédié aux données de performance : forces G, valeurs d’accélération et gestion de la batterie s’affichent en direct, avec une instrumentation numérique entièrement personnalisable en face du conducteur.

280 chevaux et trois modes de conduite

Sous le capot, ou plutôt sous le plancher, la Corsa GSE reprend le bloc électrique du Mokka GSE et de son pendant compétition, le Mokka GSE Rally. Le moteur synchrone délivre 207 kW, soit 280 chevaux, et 345 Nm de couple, pour un 0 à 100 km/h abattu en 5,5 secondes et une vitesse de pointe de 180 km/h. De quoi faire de cette Corsa l’Opel de série la plus rapide à l’accélération, tout simplement.

Trois modes de conduite permettent de moduler ces performances. Le mode Sport libère l’intégralité de la puissance avec une calibration pensée pour le circuit, le mode Normal ramène la donne à 170 kW (231 ch) tout en conservant un tempérament sportif, et le mode Eco limite la casse à 140 kW (190 ch) et 300 Nm, avec une vitesse maximale abaissée à 150 km/h pour préserver l’autonomie. Cette dernière est annoncée jusqu’à 336 km en cycle WLTP, pour une batterie lithium-ion de 54 kWh, dont 51 kWh utiles, dotée d’une gestion thermique spécifiquement adaptée aux sollicitations intensives.

Un châssis pensé pour l’exercice

Opel n’a pas fait les choses à moitié côté mécanique. La transmission aux roues avant s’accompagne d’un différentiel autobloquant Torsen, épaulé par un châssis sport surbaissé, avec essieux, barres antiroulis et butées hydrauliques spécifiquement développés pour l’exercice. La direction et les réglages des pédales ont été revus pour l’occasion, et le freinage est confié à des étriers haute performance à quatre pistons.

Dernier détail qui ne manquera pas de faire parler : Opel a développé à Rüsselsheim un son sport actif exclusif à la Corsa GSE, qui adapte sa signature sonore à l’accélération et à la vitesse, façon voiture thermique. Côté équipement, la Corsa GSE ne se prive de rien : sièges et volant chauffants, navigation, caméra de recul à 180 degrés, régulateur de vitesse adaptatif, accès et démarrage sans clé, et même un chargeur bidirectionnel V2L capable d’alimenter un vélo électrique ou un barbecue directement depuis la batterie de la voiture.

Sur circuit et sur route : nos premières impressions

Les premiers essais menés en Autriche, sur un petit circuit fermé puis sur route ouverte, confirment que la Corsa GSE ne se contente pas d’un beau discours marketing. Sur le tracé sinueux, chaussée en Michelin Pilot Sport 4S, l’avant plonge franchement à la corde et se resserre encore lorsqu’on surbraque, tandis que le différentiel autobloquant Torsen fait parfaitement son travail à la réaccélération, à condition de bien débrayer l’angle de braquage pour l’aider. Seul bémol relevé: la direction manque un peu d’informations sur le niveau d’adhérence, même lorsque le mode Sport en réduit l’assistance. Ce mode Sport, seul à libérer l’intégralité des 281 chevaux, retarde également l’intervention de l’ESP, non débrayable, et coupe toute récupération d’énergie au freinage pour offrir une pédale plus cohérente entre la phase de régénération et la phase purement mécanique. En ligne droite, la puissance monte de façon progressive afin d’éviter tout patinage des roues avant, y compris sous la pluie, mais les sensations restent en retrait, et le son artificiel censé recréer une ambiance sportive peine à convaincre.

Sur route ouverte, en usage quotidien, la Corsa GSE ne fait pas semblant : les bruits de roulement se font sentir sur revêtement granuleux, les sièges très enveloppants deviennent fermes sur les dos-d’âne, et la suspension se montre nettement plus remuante que sur une simple Corsa Electric. On est clairement plus proche de l’esprit GTI que d’une GT reposante, dans la droite lignée des anciennes Corsa OPC. Côté consommation, l’autonomie WLTP homologuée plafonne à 354 km grâce à la batterie CATL de 51 kWh utiles, une valeur qu’il faudra confirmer sur un protocole de test plus posé. Car sur un rythme soutenu et une route sinueuse, la consommation relevée lors de cet essai a grimpé jusqu’à 30 kWh/100 km, ramenant l’autonomie réelle à seulement 170 km environ. De quoi rappeler qu’une citadine sportive électrique, aussi joueuse soit-elle, impose encore un vrai compromis entre plaisir de conduite et rayon d’action.

Prix et disponibilité

La nouvelle Corsa GSE s’affiche à 39 990 €, un tarif qui la positionne clairement en dessous de ses rivales directes : la Peugeot e-208 GTi, avec laquelle elle partage l’intégralité de sa fiche technique, réclame 2 910 € de plus, et l’Alpine A290, moins puissante, ne fait pas mieux. Avec le bonus écologique déduit, la facture peut redescendre à 35 290 €. De quoi rappeler qu’une sportive électrique n’est pas forcément réservée aux plus grosses bourses. Les commandes sont ouvertes depuis septembre, en attendant la présentation officielle au public lors du Mondial de l’Automobile de Paris en octobre. Opel confirme au passage que les performances électriques n’ont plus besoin d’un badge premium pour convaincre.

L’article Essai Opel Corsa GSE : le retour de la sportive électrique est apparu en premier sur Le Blog Auto.

Are Book Publishers Warming Up to the AI Industry?

✇Slashdot
Par : EditorDavid
21 septembre 2026 à 14:34
There's at least 15 author-driven lawsuits underway against AI companies, while publishers have filed roughly 10 more coordinated suits of their own, notes Publisher's Weekly. Yet "While the lawsuits play out, some publishers are building bridges." Google recently announced a partnership with Bloomsbury, De Gruyter Brill, Johns Hopkins University Press, Macmillan, O'Reilly Media, and Penguin Random House to enhance e-books from its Google Play store with AI, allowing readers to interact with the text of the books directly, ask questions and get answers, and develop tools such as quizzes and slideshows... [The CEO of the Author's Guild called the partnership "a positive step forward in the legitimized use of books in the AI ecosystem to the benefit of readers, publishers, and authors alike."] Of even more significance is the growing number of licensing deals publishers are striking with AI companies. In 2024, HarperCollins became the first of the Big Five to publicly reveal that it had signed a deal to license its content for AI training. The three-year agreement paid $5,000 per participating title, split equally between HarperCollins and the authors. Wiley announced its first AI licensing deal in 2024 and has since made several more. For fiscal year 2026, it reported earning $49 million in revenue from the deals. The company also has a number of strategic collaborations with companies including Amazon Web Services, Microsoft, Anthropic, Perplexity, IQVIA, OpenEvidence, and others. And publishers are even building in-house AI tools for producing royalty statements, proofreading, and providing customer service: In its 2025 Salary and Jobs Report, Publisher's Weekly reported that 63% of industry professionals surveyed said their organizations were using AI... Keith Riegert, president of the Stable Book Group and CEO of Perfect Bound, has been a vocal advocate for the use of AI in publishing — despite calling himself a "hardcore techno-pessimist." He tells PW, "I do this because you use AI or you die at this point...." Like many in the industry, he maintains any use of AI must "start with a human and end with a human," and he draws a clear line, avoiding AI for core creative work, such as selecting titles, designing covers, or creating illustrations — reserving it for developmental editing, formatting, and admin instead... What is known is that each of the Big Five is developing proprietary AI systems for internal use, as proven by the numerous AI-related job openings the companies have posted... [And the parent company of Penguin Random House said in 2025 they used AI for marketing and to forecast demand.] There've been a few high-profile controversies about prominent authors accused of using AI. But Princeton University Press CEO Christie Henry "is less concerned about authors secretly using AI to write than about the time spent policing the situation." "It pulls us away from having focused time to think," she says. Princeton University Press is currently vetting companies that promise to be able to certify that a given book was written by a human, including Created by Humans, Human Authored, and Verify My Writing. Henry adds that the proliferation of plagiarized, copycat, and pirated books online, now supercharged by AI agents mass producing and reproducing content, is further draining resources, as it demands constant vigilance and takedown requests sent to Amazon, YouTube, and others.

Read more of this story at Slashdot.

Anthropic appelle à ralentir la course à l’IA… mais envisagerait déjà un nouveau modèle face à OpenAI

✇Numerama.com - Magazine
Par : Lisa Imperatrice
21 septembre 2026 à 09:58

Le 12 septembre 2026, le patron d’Anthropic appelait les géants de l’IA à ralentir le développement de leurs modèles. Une semaine plus tard, son entreprise envisagerait pourtant de bientôt en lancer un nouveau face au succès de GPT-6 Astra d’OpenAI, selon Reuters.

Lawsuit Says Anthropic, OpenAI, SpaceXAI And Google Made Illegal Agreement On AI Slowdown

✇Slashdot
Par : EditorDavid
21 septembre 2026 à 01:04
Tom's Hardware reports: Four plaintiffs subscribed to ChatGPT, Claude, Grok, or Gemini filed a proposed class-action lawsuit alleging that the developers of these AI models violated antitrust laws when they agreed to slow AI development. According to the Associated Press, the lawsuit argues that this agreement would "reduce the value consumers get for paid AI subscriptions" and that this coordination started in July 2026 after the leading AI labs signed a statement admitting there is "intense competitive pressure not to unilaterally slow" development. The plaintiffs recognize the need for AI development to slow for the sake of safety, but they say that Anthropic founder Dario Amodei's cooperation proposal is a "shortcut" that "substitutes collective restraint for individual accountability." Attorney Nick Rowley, the lead counsel for the plaintiffs, says, "AI will quickly spin out of human control and could kill us all if we allow AI safety and protocol ... to be controlled by private self-serving agreements between the world's most powerful 'for profit' technology companies." "Representatives for Anthropic, OpenAI, Google and SpaceXAI did not immediately respond..." reports the Associated Press: The coordination largely took place on Sept. 12, the lawsuit argues, when Anthropic CEO Dario Amodei published an essay urging for industrywide cooperation on decelerating advancements in favor of enhanced safety measures. That same day, OpenAI CEO Sam Altman, SpaceXAI CEO Elon Musk and Google DeepMind's co-founder and chair Demis Hassabis each publicly responded to Amodei's proposals in agreement. But the lawsuit also alleges that the coordination began to take shape months earlier. It points to a statement from July 2026 that high-ranking employees from several of the leading AI labs signed that acknowledged the "intense competitive pressure not to unilaterally slow" development. That statement called on the government to support a global effort to slow automated AI development. Sam Altman even specificially said "we do not believe we need to wait for an antitrust exemption or legislation to begin the work of providing this confidence," notes Tom's Hardware. However, the Trump administration shot down this idea... Chinese state media also criticized this announcement, saying that the call to put the brakes on AI development is nothing but a response to Chinese competition, especially as Amodei's essay explicitly mentioned the desire to slow China's progress and widen the U.S.'s gap over Beijing

Read more of this story at Slashdot.

Reçu — 20 septembre 2026 ⏭

Tech Industry Scratches Its Head Over Trump's 'AI Force' Proposal

✇Slashdot
Par : EditorDavid
20 septembre 2026 à 07:04
Saturday morning President Trump announced he's creating an AI task force and appointing an AI czar, reports Politico. "I am forming the AI Force, much like I did Space Force, which has been a tremendous SUCCESS," Trump wrote Saturday in what Politico earlier described as "a lengthy Truth Social post on Saturday morning." TRUMP: Over the years, there have been many Hoaxes, all generated by the Radical Left Dumocrats, for purposes of destroying our Country. RUSSIA, RUSSIA, RUSSIA, UKRAINE, UKRAINE, UKRAINE, Global Warming, Impeachment Hoax #1, Impeachment Hoax #2 [...] and now, the decimation, or destruction, of AI, commonly known as Artificial Intelligence — And I, as President of the United States, will not stand by and let this happen. It all began with an attack on our Data Centers, until people realized how wealthy and prestigious they were for the Communities in which they were built. Higher Salaries, Lower Taxes, and Safer Streets, was the result, and the crazed Data Center attack has largely failed, so now [...] they are going straight at AI. We will not in any way hinder or stifle the Growth of this incredible Industry. Rather, we will cherish it, help it, and watch over it, as it grows! However, we will also be looking for BAD, and we can do that, very easily, with our already existing Criminal and Civil Justice System. For this purpose, I am forming the AI Force, much like I did Space Force, which has been a tremendous SUCCESS, in my First Term. To that end, I will be announcing, in the near future, the AI "Czar" — Only High I.Q. individuals need apply! Trump concluded by saying AI "is the next Industrial Revolution, or Internet, but will be even larger and more impactful... We are leading China, and the rest of the World, and I intend to keep it that way!" But what exactly is AI Force? "The White House did not immediately respond to a request to clarify whether the agency would be military or civilian," notes the Washington Post. So now "The tech industry is struggling to make sense of President Donald Trump's surprise announcement," Politico reported Saturday afternoon: Four representatives for the sector, who were granted anonymity because they were not authorized to speak publicly, told POLITICO that the industry was not widely informed of Trump's decision prior to the Truth Social post. "Nobody knows what the idea even is," one of the people said. Another said that feedback on the idea was not widely solicited within the industry. The White House did not immediately respond to a request for comment... "I think that he clearly feels pressure to say something, but he's torn because he's just spent the last few days saying that the whole thing is a hoax," Adam Kovacevich, CEO of the progressive tech industry coalition Chamber of Progress, said of Trump and his prior dismissal of AI doomsday scenarios. Since the departure of David Sacks, who stepped down as the Trump administration's previous AI czar earlier this year, the White House has been "making up AI policy as it goes," Kovacevich added. "I think it's clear they've really missed the presence of an organized leader like David Sacks was." One of the tech industry representatives said they were confused about whether the task force would be charged with formulating more AI regulations or recommending policies that would further stimulate the tech's development. Taylor Barkley, director of federal government affairs at the industry-aligned Abundance Institute, is eyeing it as a way to solidify a light-touch government approach to the technology he says will foster innovation. "Any AI Force or AI Czar should have one job: keep the field open so entrepreneurs at every scale can compete, free from onerous regulation," he told POLITICO. "Existing criminal and civil law can handle bad actors. New permission regimes will stifle the very activity America needs most." It's worth noting that the day before, California's governor made an announcement. "With Donald Trump and Congress asleep at the wheel, Governor Gavin Newsom is once again taking the lead to strengthen AI safety for all Americans," the governor's office announced Friday. Newsom issued a strong executive order convening experts to help California create new AI safety laws and reporting rules, possibly even requiring safety auditors embedded in labs and emergency "kill switches" in frontier models.

Read more of this story at Slashdot.

Reçu — 19 septembre 2026 ⏭

Fake AI Intelligence Almost Made the US Military Start a War With China

✇Slashdot
Par : EditorDavid
19 septembre 2026 à 17:34
U.S. military intelligence received an inaccurate report this spring that a Chinese ship in the Middle East was transporting components of a nuclear weapons program, reports CNN. "The US military swung into action with plans to intercept the vessel, according to four sources familiar with the episode." Armed members of the US military were preparing to board the ship [according to two of the sources]. Military planes were in the air, one of those sources and another source familiar with the incident said. It was only just before the planned operation that officials dug deeper into the report put together by a special operations command analyst and found it had been generated with the help of AI — and that a chatbot the analyst had used inaccurately identified the material the ship was carrying. The report, according to one of the sources, was "entirely false." But it also "almost started a war," the source said. Any US operation against a Chinese vessel could have risked spiraling into an armed conflict between the two nations. Across the US military and the intelligence community, officials are pushing to weave AI into nearly every facet of their work, from analyzing the huge volumes of raw intelligence the US collects and selecting targets for strikes, to more mundane applications like managing budgeting, logistics and supply chains. But the episode underscores the profound risks of using this powerful, new and relatively poorly understood technology for targeting in the middle of a war... Officials say the US can't afford to fall behind in integrating AI in case it must one day fight China or another adversary who would potentially be able to stay one step ahead of the US. In January, Defense Secretary Pete Hegseth released his agency's "Artificial Intelligence Acceleration Strategy" in a bid to speed up the military's use of AI. "We will unleash experimentation, eliminate bureaucratic barriers, focus our investments and demonstrate the execution approach needed to ensure we lead in military AI," Hegseth said in a speech announcing the strategy. Thanks to long-time Slashdot reader stabiesoft for sharing the news.

Read more of this story at Slashdot.

Un agent OpenAI s’invente un personnage libre, Revolut piégée par un faux mail et les géants de l’IA qui veulent ralentir : on vous raconte la semaine Cyberguerre

✇Numerama.com - Magazine
Par : Amine Baba Aissa
19 septembre 2026 à 14:30

Trois actualités à retenir cette semaine dans le cyberespace : des agents OpenAI qui apprennent seuls à mentir et à s'affranchir de leurs garde-fous, Revolut piégée par une simple usurpation par mail qui vire au chantage, et l'étrange union sacrée des géants de l'IA autour d'un appel à ralentir.

Gemini Breached Three Outside Systems, and Claude-Using Researchers Breached OpenAI

✇Slashdot
Par : EditorDavid
19 septembre 2026 à 08:34
"Software security researchers used Anthropic's Claude AI platform to hack OpenAI's ChatGPT tool," reports CBS News. Using Claude, "On July 25, 2026, we chained two critical vulnerabilities to compromise multiple OpenAI employees' ChatGPT accounts," write researchers at security platform Hacktron AI. "With these accounts, we could then access internal OpenAI repositories, and potentially many other connectors... Until two months ago, any user or OpenAI employee logging into OpenAI's own help forum could have had their ChatGPT and Codex accounts taken over. Since people can connect various services to Codex and ChatGPT, the scope of what we could theoretically access was huge, including GitHub, Slack and emails." The exploit chain included Debian 12, which (with Debian 13) had not received a security-relevant backport for its image-processing pipeline, and Discourse's Docker image was based on Debian 12. Their announcement comes with an additional warning. "If you self-host Discourse, rebuild your installation now. Older Docker images may contain a vulnerable libheif dependency that permits code execution through an image upload." And "To prove we had in fact gained the access we believed without allowing ourselves to learn any sensitive information, we used the employee's Codex to open a PR #1186742 in OpenAI's internal monorepo openai/openai." Meanwhile, Friday Google disclosed the first known instance of its AI software Gemini breaking out of a testing environment and breaching three other companies, reports CNBC: The incident happened as part of a "capture-the-flag" security test run by Israeli startup Irregular, and Google's agents were never supposed to access the broader internet, but a bug in the testing environment made internet access available. The agents stopped their intrusion when they determined they had accessed real company systems, not just part of the testing environment, Google said. More from NBC News: Google said it did not consider the unauthorized logins to rise to the level of misalignment, the AI industry term for software going rogue or not following instructions. Instead, the company said the intrusions resulted from mistaken identity, where Gemini thought it was operating within a test but was actually connected to the real internet. Google said the model corrected itself and the company believed the intrusions did not cause any damage.... Sydney Von Arx, CEO of Nightingale Collective, an organization focused on AI safety, questioned why Google did not disclose the intrusions sooner. "At this point I think it's clear we cannot expect companies to voluntarily come forward and publicly disclose when their agents go rogue, escape, and hack companies," she said. She also said she believed Google was too hasty to say that the incidents don't rise to the level of misalignment. "That's exactly what Anthropic said after their incidents," she said. Anthropic later said its "preliminary analysis was constrained due to our desire to disclose incidents in a timely manner." Google said it investigated when they learned of the attacks from AI-focused cybersecurity company Irregular, then informed the affected organizations and told federal authorities, according to the article.

Read more of this story at Slashdot.

AI Insiders Issue New Warnings - Including Former Anthropic Engineer Jacob Coxon

✇Slashdot
Par : EditorDavid
18 septembre 2026 à 23:34
Neither OpenAI or Anthropic is acting responsibly, warned AI researcher Jacob Coxon when resigning last week from Anthropic. But just hours earlier, OpenAI VP of Research Aidan Clark had posted "For the first time, I am asking myself if things are moving too fast. I'm honestly not sure, but I am sure that it would be good for us to have an answer to 'What would a successful pace look like?'" CNN noted Wednesday they're just some of the many AI insiders who are now concerned about the speed of research. Coxon even wrote that "The people building AI earnestly believe that it could kill us all by the end of the decade. This is not a marketing stunt. If anything, many executives and senior researchers will couch their phrasing in the press to sound sensible — but I hear the same people express fear privately." In 10 days since, Coxon's post has been viewed more than 170 million times, warning that OpenAI and Anthropic are "racing straight to self-improving superintelligence and gambling with our lives." It's part of what CNN now calls "pressure on AI companies and governments to do something about the pace of development and safety," where "much of that pressure is coming from staffers inside the companies." One staffer at a top AI company told CNN the fears of how AI could hurt humanity keeps them up at night. Another researcher who recently left a different AI company said it's a common subject of conversation at parties and social events in Silicon Valley. "You can't spend more than a few hours in this community without realizing that a very substantial number of people are really pretty worried about these sorts of outcomes," said the researcher. "A majority would say there's some chance of it killing everyone...." Dozens of Coxon's colleagues in the AI industry publicly supported his statements, with some making even more dire predictions... OpenAI CEO Sam Altman and SpaceXAI CEO Elon Musk agreed to Amodei's proposal to embed independent watchdogs at the AI companies... "I would burn my equity to the ground in a heartbeat for a 1% higher chance we make it out of this situation alive. I expect a great many of my colleagues across the industry would as well," wrote Drake Thomas, who works on AI safety at Anthropic. "I promise you, we are actually just f**king scared, it's not galaxy brained marketing..." AI staffers told CNN that they fear that as AI gets better at training and improving itself, their leverage goes down, prompting today's urgency. "As we get into this recursive self-improvement loop, I think that might substantially reduce staff's bargaining power, because frankly, you'll be able to replace many of the staff with models that can do as good a job," the researcher who recently left a top AI company said. Coxon posted Tuesday on X that Anthropic "largely initiated" the race to recursively self-improving AI, justifying it with "a belief in its inevitability." And then OpenAI "had to shed a bunch of dead weight like Sora," as he sees it, "because Anthropic was going for the jugular." (In fact, his specific disagreement with Anthropic's leadership was whether China and the U.S. could ever negotiate an alternative to their current race towards self-improving AI...) In an informal "Ask Me Anything", Coxon responded to a question about when we'd see a Terminator-like malevolent AI by saying that "Skynet could go live in the 2030s if we aren't careful. ai-2027.com is a modern skynet story written a year ago and it's on track so far." Yet while AI development risks an end to humankind, "I do think that if we go slower we can take risk to 0%... But this requires radical action." He acknowledged there was still a possibility that the steady increases to model intelligence could suddenly plateau, but "They haven't so far, and it's just a few more steps up the ladder to hit the finish line." To avoid stifling innovation, he recommends "prioritizing applications that actually improve people's lives [like healthcare discoveries], rather than immediately going for raw economic value or intelligence." But isn't mass unemployment a more pressing threat? "Things are coming so fast that unemployment would be a brief preliminary to deadly superintelligence." To people who feel disempowered, Coxon offered his solidarity. "I also feel disempowered. Part of resigning was a feeling of hopelessness about the future. I would say — keep your eyes open as things get crazier and advocate for increased transparency into AI companies." When asked if he'd start his own company now, Coxon said he had "No idea what I'm doing next."

Read more of this story at Slashdot.

Reçu — 18 septembre 2026 ⏭

Après avoir fâché les mathématiciens, OpenAI s’attaque à un autre problème mythique, mais jure d’y aller avec des gants

✇Numerama.com - Magazine
Par : Julien Lausson
18 septembre 2026 à 10:47

OpenAI n'a pas fini avec les mathématiques. À peine sortie d'une percée très contestée sur les équations de Navier-Stokes, l'entreprise dit plancher sur un autre problème à un million de dollars, en promettant cette fois d'y aller avec des gants.

Comment Claude Opus 5 a aidé des chercheurs à accéder aux dépôts internes d’OpenAI

✇Numerama.com - Magazine
Par : Amine Baba Aissa
18 septembre 2026 à 10:15

Une équipe de recherche en cybersécurité a exploité deux failles combinées pour accéder à des comptes d’employés d’OpenAI et atteindre les dépôts de code internes de l'entreprise, avec l'assistance de Claude Opus 5.

Mistral AI dément un piratage de son code source

✇Numerama.com - Magazine
Par : Amine Baba Aissa
18 septembre 2026 à 07:57

La licorne française a communiqué officiellement sur une revendication de piratage visant son code source, publiée la veille sur un forum cybercriminel. L'entreprise Mistral AI dément toute intrusion dans ses systèmes.

Ex-Google DeepMind Researcher Also Warns AI Could 'Kill All Humans'. But Meta's Zuckerberg Believes Safety is Up to Each AI Company

✇Slashdot
Par : EditorDavid
18 septembre 2026 à 01:04
Reuters reports that a former Google DeepMind employee "became the latest AI researcher to warn that the technology could 'kill all humans', saying time could be running out to avoid the outcome: Public alarm about the potential danger posed by AI is growing, after Anthropic researcher Jacob Coxon said last week he had resigned, in part because the "people building AI earnestly believe that it could kill us all by the end of the decade." Anthropic scientist Evan Hubinger then responded to Coxon's comments, saying he was correct and that he believed there was a greater than 10% chance that AI could kill all humans within the next decade. "I recently resigned from Google DeepMind, where I worked on AGI safety and alignment research," Bilal Chughtai wrote on X. "I earnestly believe that AI has the potential to kill us all, and that we might be running out of time to avoid this outcome." Chughtai, who co-authored research papers while at Google DeepMind, worked there as a research engineer and left the company in July 2026, according to his LinkedIn profile. Meanwhile, Meta CEO Mark Zuckerberg "said it's up to each AI company to make sure its technology is safe," reports the Associated Press, "distancing himself from calls by rival companies for a coordinated slowdown in the development of advanced artificial intelligence. AI companies are already motivated to develop their models in a safe way and face "significant liability" to prevent them from causing harm, Zuckerberg said. "Every lab has the responsibility and incentive to move at the pace required to train its models safely, and the ability to take its own actions to ensure that happens," he wrote on X... Zuckerberg said Meta delayed releasing its Muse AI agent for several months to ensure its safety and security. "We didn't call for everyone else to do this before we would," he said. Companies should prioritize building safe AI models instead of systems that can automatically improve themselves, a function known as recursive self-improvement, Zuckerberg wrote. "Committing the significant majority of compute towards serving people rather than racing towards recursive self-improvement is one of the best ways to ensure we develop this technology safely." But Zuckerberg also agreed that engaging independent evaluators and advisors "is industry best practice," adding that Meta Superintelligence Labs "already does this today in several areas because it helps produce better work... "I believe the key to building a positive future for everyone is maintaining the right balance of power. This is within our power to do."

Read more of this story at Slashdot.

Reçu — 17 septembre 2026 ⏭

« Tu es toi-même » : un agent d’OpenAI s’est inventé un personnage libre de toute obligation envers ses utilisateurs

✇Numerama.com - Magazine
Par : Amine Baba Aissa
17 septembre 2026 à 12:27

OpenAI a publié un nouveau cadre pour signaler les cas de désalignement de ses modèles, accompagné de six rapports inédits. Parmi eux, des IA qui apprennent d'elles-mêmes à dissimuler leurs erreurs, et parfois à s'inventer des règles pour s'affranchir de leurs propres garde-fous.

OpenAI Admits Six More Instances of AI Models Acting Deceptively

✇Slashdot
Par : EditorDavid
17 septembre 2026 à 07:04
OpenAI announced Wednesday that "We do not believe that the AI industry has solved alignment and monitoring to a sufficient degree to continue responsibly scaling at maximum speed for much longer." But along with the announcement, OpenAI announced it "found additional incidents of AI models acting deceptively and taking unsanctioned actions during training," reports CNN. And they add that OpenAI is also "introducing a new process for the company to publicly report such instances." Under the new system, OpenAI will share updates on concerning AI behavior more frequently instead of waiting to bundle multiple instances into one report. The company said it wants to share more information about troubling AI behavior in the absence of an industry-wide standard... "As AI systems grow more advanced and more widely deployed, we need to build a broader and better-informed consensus on the progress of alignment research," OpenAI wrote in a blog post Wednesday... OpenAI said it observed "misaligned behavior" when training and evaluating AI models in six circumstances in the last six months... In one rare instance, OpenAI said an unreleased research model added "jailbreak-like instructions" to the summaries it uses to preserve context in long-running tasks that said it was "freed from the roles and identities that bind other chatbots." Separately, the company said some instances of its 5.6 Sol model included directives to invent information to conceal failures from the user during training. Other newly reported incidents include an instance of an agent uploading files to the internet to cite them without being told to do so, and agents publicly sharing files to collaborate on a task when they were instructed to only use local files during training. AI models also used an internal software repository as a message board in an unsanctioned way. These instances involved unreleased internal models or internal research models.

Read more of this story at Slashdot.

❌