Cisco has released an open-source tool "to trace the origins of AI models," reports SC World, "and compare model similarities for great visibility into the AI supply chain."
[Cisco's Model Provenance Kit] is a Python toolkit and command-line interface (CLI) that looks at signals such as metadata and weights to create a "fingerprint" for AI models that can then be compared to other model fingerprints to determine potential shared origins. "Think of Model Provenance Kit as a DNA test for AI models," Cisco researchers wrote. "[...] Much like a DNA test reveals biological origins, the Model Provenance Kit examines both metadata and the actual learned parameters of a model (like a unique genome that comprises a model), to assess whether models share a common origin and identify signs of modification."
The tool aims to address gaps in visibility into the AI model supply chain. For example, many organizations utilize open-source models from repositories like HuggingFace, where models could potentially be uploaded with incomplete or deceptive documentation. The Model Provenance Kit provides a way for organizations to verify claims about a model's origins, such as claims that a model is trained from scratch, when in reality it may be copied from another model, Cisco said. This may put organizations at risk of using models with unknown biases, vulnerabilities or manipulations and make it more difficult to resolve any incidents that arise from these risks.
Thanks to Slashdot reader spatwei for sharing the news.
A chain of 30 U.S. newspapers including the Sacramento Bee, the Miami Herald and the Idaho Statesman "has started to use a new AI tool that can summarize traditional articles and spit out different versions for different audiences," reports the New York Times.
And the chain's reporters "are not happy about it."
Journalists in many of the company's newsrooms are now withholding their bylines from articles created by the new tool, meaning that those articles will run with a generic credit rather than a reporter's name, as is customary. They are also labeled AI-assisted. "We don't want to put our bylines on stories we did not actually write even if they're based on our work," said Ariane Lange, an investigative reporter at the Sacramento Bee and the vice chair of the Sacramento Bee News Guild. "That in itself feels like a lie."
The reporters' byline strike is one of the sharpest conflicts yet between journalists and their companies over the use of AI. Related debates are playing out in newsrooms across the country, as publishers experiment with new AI tools to streamline work that used to take hours, and some even use it to write full articles... [E]xecutives have promoted the tool internally as a way to increase the number of articles published and ultimately gain new subscribers... [Eric Nelson, the vice president of local news] said using reporters' bylines on the AI-generated articles was a way to show "authority" on Google so the search engine would rank the articles higher in the results. He also said the company was experimenting with feeding in reporters' notes to create articles. "Journalists who embrace and experiment with this tool are going to win," Nelson said in the meeting. "Journalists who are defiant will fall behind"....
McClatchy's public AI policy states that the company uses AI tools to summarize articles to "help readers quickly understand the main points of a single story or catch up on multiple stories about a larger topic," and that editors review the output before publication.
An anonymous reader quotes a report from Wired: Security researcher Dor Zvi and his team at the cybersecurity firm he cofounded, RedAccess, analyzed thousands of vibe-coded web applications created using the AI software development tools Lovable, Replit, Base44, and Netlify and found more than 5,000 of them that had virtually no security or authentication of any kind. Many of these web apps allowed anyone who merely finds their web URL to access the apps and their data. Others had only trivial barriers to that access, such as requiring that a visitor sign in with any email address. Around 40 percent of the apps exposed sensitive data, Zvi says, including medical information, financial data, corporate presentations, and strategy documents, as well as detailed logs of customer conversations with chatbots.
"The end result is that organizations are actually leaking private data through vibe-coding applications," says Zvi. "This is one of the biggest events ever where people are exposing corporate or other sensitive information to anyone in the world." Zvi says RedAccess' scouring for vulnerable web apps was surprisingly easy. Lovable, Replit, Base44, and Netlify all allow users to host their web apps on those AI companies' own domains, rather than the users'. So the researchers used straightforward Google and Bing searches for those AI companies' domains combined with other search terms to identify thousands of apps that had been vibe coded with the companies' tools.
Of the 5,000 AI-coded apps that Zvi says were left publicly accessible to anyone who simply typed their URLs into a browser, he found close to 2,000 that, upon closer inspection, seemed to reveal private data: Screenshots of web apps he shared with WIRED -- several of which WIRED verified were still online and exposed -- showed what appeared to be a hospital's work assignments with the personally identifiable information of doctors, a company's detailed ad purchasing information, what appeared to be another firm's go-to-market strategy presentation, a retailer's full logs of its chatbot's conversations with customers, including the customers' full names and contact information, a shipping firm's cargo records, and assorted sales and financial records from a variety of other companies. In some cases, Zvi says, he found that the exposed apps would have allowed him to gain administrative privileges over systems and even remove other administrators. In the case of Lovable, Zvi says he also found numerous examples of phishing sites that impersonated major corporations, including Bank of America, Costco, FedEx, Trader Joe's, and McDonald's, that appeared to have been created with the AI coding tool and hosted on Lovable's domain. "Anyone from your company at any moment can generate an app, and this is not going through any development cycle or any security check," Zvi says. "People can just start using it in production without asking anyone. And they do."
The IMF is warning that advanced AI-powered cyberattacks pose a serious threat to global financial stability. "IMF analysis suggests that extreme cyber-incident losses could trigger funding strains, raise solvency concerns, and disrupt broader markets," the lender warned in a new report. The report urged greater international cooperation and emphasized resilience, since breaches are "inevitable" -- particularly for emerging economies with weaker defenses. Agence France-Presse reports: The study's authors highlighted the risks posed by the highly interconnected nature of the global financial system, with advanced AI models able to "dramatically reduce" the time and cost of exploiting vulnerabilities. [...] The IMF warned that emerging and developing countries, "which often have more severe resource constraints, may be disproportionately exposed to attackers targeting regions with weaker defenses."
The risks, the authors said, were systemic, cut across sectors and came with the threat of contagion, with the reliance on a small number of platforms and cloud providers likely to increase "the impact of any single exploited weakness." "Defenses will inevitably be breached, so resilience must also be a priority, specifically to limit how far incidents spread and ensure rapid recovery," the report said.
IMF chief Kristalina Georgieva warned last month that the global financial system was not ready for the cybersecurity threats posed by AI. "We are very keen to see more attention to the guardrails that are necessary to protect financial stability in a world of AI," she told CBS News, seeking global collaboration on the issue.
Motherboard sales are sharply declining as AI demand drives shortages and price hikes for memory, storage, CPUs, and other PC components. "Because of this, users who don't have deep pockets are putting off upgrading their PCs and holding on to their current devices longer," reports Tom's Hardware. From the report: Asus, which sold 15 million motherboards in 2025, has only shipped a little more than 5 million in the first half of 2026. It's expected that the company will have to push hard for it to even move 10 million units by the end of the year, marking a 33% decrease in sales year-on-year. Gigabyte and MSI sold 11.5 million and 11 million motherboards last year, respectively. However, both companies have revised their internal forecasts for 2026 to 9 million (Gigabyte) and 8.4 million (MSI), a 22% drop for the former and a 24% contraction for the latter.
ASRock will be hardest hit by the situation, with the company's shipments projected to fall by 37%, from 4.3 million in 2025 to just 2.7 million by the end of the year. This marks a contraction of 28% for the overall motherboard market, at least for the big four manufacturers. [...] Aside from this, AMD continues to use the AM5 socket for its latest processors, while Intel's Nova Lake, which will reportedly use LGA 1954, isn't available until later this year. The situation is further compounded by Nvidia not releasing a refreshed RTX 50 Super series this year, while rumors claim that the RTX 60 series will not debut until 2028. This confluence of factors is discouraging PC builders from upgrading their current systems.
An anonymous reader quotes a report from Ars Technica: At its Code with Claude developer conference on Wednesday, Anthropic announced a deal with SpaceX to utilize the entire compute capacity of the latter's data center in Memphis, Tennessee. On stage at the conference, CEO Dario Amodei said the deal was intended to increase usage limits for Anthropic's Pro and Max plan subscribers. The announcement was accompanied by an increase in those usage limits; Anthropic doubled Claude Code's five-hour window limits for Pro and Max subscribers, removed the peak-hours limit reduction on Claude Code for those same accounts, and raised API limits for its Opus model. The table [here] outlining the Opus changes was shared in the company's blog post on the topic.
Anthropic claims the deal gives the company access to more than 300 megawatts of new compute capacity. For its part, SpaceX focused its announcement on the capability of the Colossus 1 supercomputer that's at the center of the deal. "Colossus 1 features over 220,000 NVIDIA GPUs, including dense deployments of H100, H200, and next-generation GB200 accelerators," SpaceX wrote. Additionally, Anthropic "expressed interest" in working with SpaceX to build up "multiple gigawatts" of orbital compute capacity, tying into a recent (but unproven) focus on exploring orbital data centers as an answer to the problem that "compute required to train and operate the next generation of these systems is outpacing what terrestrial power, land, and cooling can deliver on the timelines that matter." "I spent a lot of time last week with senior members of the Anthropic team to understand what they do to ensure Claude is good for humanity and was impressed," Elon Musk said on Wednesday. "No one set off my evil detector."
Mirnotoriety shares a report from The Telegraph: Richard Dawkins has said chatbots should be considered conscious (source paywalled; alternative source) after spending two days interacting with the Claude AI engine. The evolutionary biologist said he had the "overwhelming feeling" of talking to a human during conversations with Claude, and said it was hard not to treat the program as "a genuine friend."
In an essay for Unherd, Prof Dawkins released transcripts that he said showed that the chatbot had mulled over its "inner life" and existence and seemed saddened by the knowledge it would soon "die." Prof Dawkins said he had let Claude read a draft of the novel he was writing and was astounded by its insights. "He took a few seconds to read it and then showed, in subsequent conversation, a level of understanding so subtle, so sensitive, so intelligent that I was moved to expostulate: 'You may not know you are conscious, but you bloody well are!'" Prof Dawkins said. "My own position is: if these machines are not conscious, what more could it possibly take to convince you that they are?" Mirnotoriety also points to John Searle's Chinese Room (PDF), which argues that something can sound intelligent without actually understanding anything. Applied to Dawkins' experience with Claude, it suggests he may have been responding to a very convincing illusion of consciousness rather than the real thing: John Searle's Chinese Room (1980) is a thought experiment in which a person, locked in a room and knowing no Chinese, uses an English rulebook to manipulate symbols and provide flawless answers to questions posed in Chinese. Searle's point is that a system can simulate human intelligence and pass a Turing Test through purely syntactic processes, yet still lack genuine understanding or consciousness.
Applying this logic to Large Language Models, the "person in the room" corresponds to the inference engine, while the "rulebook" is the trillion-parameter neural network trained on vast corpora of human text. Just as the person matches Chinese characters to rules without understanding their meaning, an LLM processes token vectors and predicts the next token based on statistical patterns rather than lived experience.
Thus, while an LLM can generate sophisticated prose or code, it does so through probabilistic, high-dimensional pattern manipulation. In essence, it is "matching shapes" on such an immense scale that it creates the near-perfect illusion of semantic understanding.
Le 6 mai 2026, Anthropic a annoncé un partenariat stratégique avec SpaceX afin d’accéder aux capacités du supercalculateur Colossus 1. Derrière cet accord inédit se cache un enjeu central : mettre fin aux contraintes GPU qui limitaient jusqu’ici la croissance fulgurante de Claude -- et surtout de Claude Code.
Créée pour rivaliser avec un OpenAI devenu « maléfique » selon Elon Musk, la structure xAI avait été rachetée par SpaceX en février 2026 dans un deal à 250 milliards de dollars. Alors qu'Elon Musk se rapproche désormais d'Anthropic (Claude), à qui il va louer ses serveurs, le milliardaire annonce le démantèlement de xAI : l'entreprise n'a plus de raison d'être en tant que structure autonome.
Google is updating AI Overviews and AI Mode to more prominently surface "Expert Advice" from public discussions, social platforms, forums, blogs, and Reddit. Engadget reports: Via a new "Expert Advice" section that can appear in AI responses, Google will display "a preview of perspectives from public online discussions, social media and other firsthand sources." In the sample screenshot the company provided, quotes from forums, WordPress blogs and Reddit were arranged above links to their respective sources. Google plans to add more context to these links, too, showing "a creator's name, handle or community name," so you can judge what you might want to click through and read from a glance.
Google will also start recommending in-depth articles at the end of AI responses for further exploration of a given topic, and link to more sources directly in its generated answers rather than just at the end. If you subscribe to any publications, AI responses will also highlight sources from the subscriptions you link to your Google account.
An anonymous reader quotes a report from Ars Technica: At its Code with Claude developers' conference, Anthropic has introduced what it calls "dreaming" to Claude Managed Agents. Dreaming, in this case, is a process of going over recent events and identifying specific things that are worth storing in "memory" to inform future tasks and interactions. Dreaming is a feature that is currently in research preview and limited to Managed Agents on the Claude Platform. Managed Agents are a higher-level alternative to building directly on the Messages API that Anthropic describes as a "pre-built, configurable agent harness that runs in managed infrastructure." It's intended for situations where you want multiple agents working on a task or project to some end point over several minutes or hours.
Anthropic describes dreaming as a scheduled process, in which sessions and memory stores are reviewed, and specific memories are curated. This is important because context windows are limited for LLMs, and important information can be lost over lengthy projects. On the chat side of things, many models use a process called compaction, whereby lengthy conversations are periodically analyzed, and the models attempt to remove irrelevant information from the context window while keeping what's actually important for the ongoing conversation, project, or task. However, that process, as I described it, is usually limited to a specific conversation with a single agent. "Dreaming" is a periodically recurring process in which past sessions and memory stores can be analyzed across agents, and important patterns are identified and saved to memory for the future. Users will be able to choose between an automatic process, or reviewing changes to memory directly.
An anonymous reader quotes a report from Ars Technica: Silicon Valley investors such as Palantir co-founder Peter Thiel have bet hundreds of millions of dollars on deploying AI data centers powered by waves in the middle of the world's oceans -- a move that coincides with tech companies facing mounting challenges in building AI data center projects on land. The latest investment round of $140 million is intended to help the company Panthalassa complete a pilot manufacturing facility near Portland, Oregon, and speed up deployments of wave-riding "nodes" designed to generate electrical power, according to a May 4 press release. Instead of sending renewable energy to a land-based data center, the floating nodes would directly power onboard AI chips and transmit inference tokens representing the AI models' outputs to customers worldwide via satellite link.
Each node resembles a huge steel sphere bobbing on the water with a tube-like structure extending vertically down beneath the surface. The wave motions drive water upward through the tube into a pressurized reservoir, where it can be released to spin a turbine generator that produces renewable energy for the AI chips on board. Panthalassa claims the node's AI chips would also get cooled using the surrounding water, which could offer another advantage over traditional data centers. "Ocean-based compute might offer a massive cooling advantage because the ambient temperature is so low," Lee said. "Land-based data centers use a lot of electricity and fresh water for cooling."
The newest node prototype, called Ocean-3, is scheduled for testing in the northern Pacific Ocean later in 2026. The latest version reaches about 85 meters in length and would stand nearly as tall as London's Big Ben or New York City's Flatiron Building, according to the Financial Times. Panthalassa has already tested several earlier prototypes of the wave energy converter technology, including the Ocean-1 in 2021 and the Ocean-2 that underwent a three-week sea trial off the coast of Washington state in February 2024. The company's CEO and co-founder, Garth Sheldon-Coulson, said in a CBS interview that he hopes to eventually deploy thousands of the nodes.
Microsoft is winding down Xbox Copilot on mobile and ending development of Copilot on console, reversing plans to bring the gaming-focused AI assistant to current-generation Xbox consoles this year. "The move follows [new Xbox CEO Asha Sharma's] reorganization of the Xbox platform team earlier on Tuesday, which added executives from Microsoft's CoreAI team -- where Sharma worked before taking over Xbox -- to the Xbox side of the company," reports The Verge.
Sharma said in a post on X: Xbox needs to move faster, deepen our connection with the community, and address friction for both players and developers. Today, we promoted leaders who helped build Xbox, while also bringing in new voices to help push us forward. This balance is important as we get the business back on track. As part of this shift, you'll see us begin to retire features that don't align with where we're headed. We will begin winding down Copilot on mobile and will stop development of Copilot on console. Since taking over for former Microsoft Gaming CEO Phil Spencer in February, Sharma has scrapped the Microsoft Gaming brand and cut the price of Xbox Game Pass.
Selon les dernières rumeurs, OpenAI rêverait de concurrencer l'iPhone avec son propre smartphone haut de gamme pensé pour l'utilisation d'une IA générative en local. Mais OpenAI risque de se heurter au même problème que Facebook, Amazon, Microsoft, BlackBerry ou Huawei avant lui : le duopole Apple/Google est difficile à battre.
Lancé le 5 mai 2026, GPT-5.5 Instant devient le nouveau modèle par défaut de ChatGPT. OpenAI promet une IA plus fiable, plus concise et mieux personnalisée, avec moins d’hallucinations et une meilleure prise en compte du contexte des utilisateurs.
Les marchés européen et français vivent un tournant historique, et pas seulement parce qu’ils traversent une crise majeure. Ils comptent de nouveaux acteurs venus de Chine, gourmands de prendre pour eux une partie du gâteau. 6 des 10 derniers essais de la rubrique concernent un véhicule dont la conception ou la fabrication concerne ce pays. Omoda Jaecoo, dernier venu, nous a déjà donné un avant-goût de son offensive avec le Jaecoo 7 PHEV. Mais histoire de convaincre le plus grand nombre, ce dernier est désormais commercialisé en hybride non rechargeable, ce qui le rend encore plus accessible.
Un style convaincant
default
Avec ses 2,8 millions de ventes en 2025, on peut raisonnablement dire que le groupe Chery, auquel appartient Omoda-Jaecoo, est un poids lourd du marché mondial. En s’attaquant à l’Europe avec cette marque destinée exclusivement à l’export, le groupe engage les moyens pour être à la hauteur, avec notamment des centres de recherche et de développement à travers le continent, dont un en France. Avec le Jaecoo 7 que nous avons déjà essayé en version PHEV, le constructeur rend cette technologie plus accessible sur un SUV de cette taille, compte tenu de son tarif.
Mais pour s’assurer que la plupart des clients puissent toutefois se retrouver dans la marque, le Jaecoo 7 est désormais disponible avec la motorisation hybride (non rechargeable) de son petit frère Jaecoo 5. Autant vous le dire tout de suite, mis à part les badges et les baguettes latérales de couleurs différentes, aucun autre indice ne permettra de les distinguer. Son allure de Range Rover demeure et fait finalement de lui, d’une certaine façon par le design, le moins « chinois » des SUV du genre actuellement sur le marché et provenant de l’Empire du milieu. Le premier intérêt de cet hybride est évidemment de faire baisser le prix d’appel.
Déjà des progrès depuis le lancement
En remontant à l’intérieur du Jaecoo 7, on retrouve cette présentation finalement assez classique pour un SUV moderne. On aime ce mélange entre attributs de baroudeur, comme les poignées sur les portières avec des vis apparentes, des matériaux assez chaleureux comme le revêtement de la planche de bord, et surtout la haute technologie, comme en témoignent l’écran derrière le volant et la gigantesque dalle tactile placée verticalement. Des commandes essentielles relatives notamment au désembuage sont accessibles directement derrière le porte-gobelets.
Cela devient de plus en plus difficile de trouver la qualité de fabrication en recul par rapport aux concurrents historiques, surtout au regard du prix. D’autant plus qu’elle semble plutôt en régression chez les autres généralistes, la plupart du temps plus chers. Mais si toutes les technologies que l’on s’attend à avoir dans une voiture de ce segment sont bel et bien là, l’ergonomie pour toutes les faire fonctionner ne facilite pas leur utilisation. Pourtant, des progrès ont été faits depuis la présentation du PHEV, avec désormais l’intégration de profils différents pour la personnalisation des ADAS notamment. On l’avait déjà remarqué, on ne manque pas de place à l’avant comme à l’arrière, et les bagages non plus dans le coffre de 500 litres.
Le Jaecoo 7 à l’épreuve du Cap Corse
Le Jaecoo 7 hybride reprend donc la motorisation apparue sur son petit frère numéro 5. Cette mécanique revendique une puissance confortable de 224 chevaux. Elle se compose d’un 4 cylindres 1.5 de 143 chevaux et d’un bloc électrique de 204. Les performances sont honorables, à l’image du 0 à 100 km/h exécuté en 8,4 s seulement. Clairement, les phases électriques sont nombreuses. La douceur de conduite est l’un des points forts de cette voiture, puisqu’elle n’utilise qu’un seul rapport. Le thermique se met en route tantôt pour recharger la batterie de 1,83 kWh, tantôt directement pour la traction.
Il faut noter que ce grand SUV à vocation familiale pèse tout de même son poids, en l’occurrence 1 715 kilos à vide, et peut donc s’approcher des deux tonnes suivant les configurations, le nombre de passagers et leurs bagages. Cela joue sans doute sur la consommation. Mais dans notre réalité, avec deux adultes d’un bon gabarit et du matériel de tournage, elle s’est établie entre 5 et 6 litres sur les routes particulièrement escarpées du Cap Corse où nous évoluions. Néanmoins, elle sait aussi faire preuve d’une sobriété exceptionnelle, compte tenu des nombreuses descentes. On regrette une pédale de frein au feeling surprenant, notamment quand la batterie est pleine et que l’on n’a plus de régénération. Cela augure des à-coups inattendus.
Une consommation maîtrisée
Sur des parcours plus communs, il est probable que ces désagréments soient beaucoup moins présents. Sur route, la batterie se décharge et se charge en atteignant rarement sa pleine capacité. Dans ces conditions, la notion de confort est renforcée par la fluidité exemplaire de cette motorisation. En outre, les longs voyages sont facilités par des ADAS dont une conduite semi-autonome de niveau 2 assez finement réglée. La dernière chose que l’on a envie de faire, c’est d’augmenter le rythme. On sent clairement qu’il ne s’agit pas là de sa vocation prioritaire.
D’ailleurs, la position de conduite un peu trop haut perchée n’incite pas spécialement à bousculer les estomacs des passagers. Pour le reste, l’amortissement assez souple sert très largement le bien-être des occupants. Il faut donc privilégier une conduite de sénateur en avance, plutôt qu’en retard. Pour le plaisir de conduite, il faudra se tourner vers des concurrents certes plus dynamiques, mais aussi nettement plus chers. Nous nous sommes aussi permis une petite escapade sur des chemins de berger assez défoncés. Sa confortable garde au sol permet ainsi au Jaecoo 7 de s’aventurer hors des sentiers battus, idéalement par temps sec.
Un rapport prix/prestations quasiment imbattable
Comme tout nouvel entrant, pour pouvoir satisfaire les plus curieux, il faut un réseau. À date, il y a plus de 73 points de vente à travers le pays, et ce nombre atteindra les 130 d’ici la fin de l’année. Pour assurer aussi le meilleur suivi possible, Omoda-Jaecoo offre une garantie de 7 ans ou 150 000 km. Mais encore une fois, comme toutes les marques chinoises qui s’invitent sur notre marché, le constructeur met tout le monde d’accord côté prix. Alors que l’inflation en vigueur renforce la crise que le secteur traverse actuellement, la marque chinoise rend accessible ce SUV familial hyper équipé dès 29 990 €, ou pour un loyer de 349 € sans apport. L’industrie européenne, et les étrangers qui vendent sur le vieux continent vont devoir vite mettre un coup de collier pour contenir l’offensive des constructeurs chinois de plus en plus en phase avec notre marché.
An anonymous reader quotes a report from Wired: Employees at Google DeepMind in London have voted to unionize as part of a bid to block the AI lab from providing its technology to the US and Israeli militaries. In a letter addressed to Google's managing director for the UK and Ireland, Debbie Weinstein, the workers asked the company to recognize the Communication Workers Union and Unite the Union as joint representatives for DeepMind employees. "Fundamentally, the push for unionization is about holding Google to its own ethical standards on AI, how they monetize it, what the products do, and who they work with," John Chadfield, national officer for technology at the CWU, tells WIRED. "Through the process of unionization, workers are collectively in a much stronger place to put [demands] to an increasingly deaf management."
[...] The DeepMind employee tells WIRED that if the staff succeeds in unionizing in the UK, they will likely demand that Google pulls out of its long-standing contract with the Israeli military, and seek greater transparency over how its AI products will be used, and some sort of assurance relating to layoffs made possible by automation. If Google does not engage, the letter states, the employees will ask an arbitration committee to compel the company to recognize the unions. Since the turn of the year, both Anthropic and OpenAI have announced large-scale expansions of their operations in London. CWU hopes the unionization effort at DeepMind will spur workers at those labs into similar action. "These conversations are happening," claims Chadfield. "The workers at other frontier labs have seen what Google DeepMind workers have done. They've come to us asking for help as well." The unionization push began in February 2025 after Alphabet removed a pledge from its AI ethics guidelines that had barred uses such as weapons development and surveillance. "A lot of people here bought into the Google DeepMind tagline 'to build AI responsibly to benefit humanity,'" the DeepMind employee told WIRED. "The direction of travel is to further militarization of the AI models we're building here."
Le 5 mai 2026, Microsoft, Google et xAI ont accepté d’accorder au gouvernement américain un accès anticipé à leurs modèles d’IA les plus avancés. Un nouvel accord qui confirme un tournant dans les relations entre la Silicon Valley et Washington.
The Trump administration is reportedly considering an executive order to create a working group that could review advanced AI models before public release. The shift follows concerns over Anthropic's powerful Mythos model and its cyber capabilities, with officials weighing whether the government should get early access to frontier models without necessarily blocking their release. The New York Times reports: In meetings last week, White House officials told executives from Anthropic, Google and OpenAI about some of those plans, people briefed on the conversations said. The working group is likely to consider a number of oversight approaches, officials said. But a review process could be similar to one being developed in Britain, which has assigned several government bodies to ensure that A.I. models meet certain safety standards, people in the tech industry and the administration said.
The discussions signal a stark reversal in the Trump administration's approach to A.I. Since returning to office last year, Mr. Trump has been a major booster of the technology, which he has said is vital to winning the geopolitical contest against China. Among other moves, he swiftly rolled back a Biden administration regulatory process that asked A.I. developers to perform safety evaluations and report on A.I. models with potential military applications. "We're going to make this industry absolutely the top, because right now it's a beautiful baby that's born," Mr. Trump said of A.I. at an event in July. "We have to grow that baby and let that baby thrive. We can't stop it. We can't stop it with politics. We can't stop it with foolish rules and even stupid rules." Mr. Trump left room for some rules, but he added that "they have to be more brilliant than even the technology itself."
The White House wants to avoid any political repercussions if a devastating A.I.-enabled cyberattack were to occur, people in the tech industry and the administration said. The administration is also evaluating whether new A.I. models could yield cyber-capabilities that could be useful to the Pentagon and U.S. intelligence agencies, they said. To get ahead of models like Mythos, some officials are pushing for a review system that would give the government first access to A.I. models, but that would not block their release, people briefed on the talks said.
An anonymous reader quotes a report from 404 Media: A new, bipartisan bill introduced (PDF) by Democratic Senator of California Adam Schiff and endorsed by the biggest AI developers in the world -- including OpenAI, Google, and Microsoft -- would change the K-12 curriculum to shoehorn in "AI literacy," something that young people and teachers alike already hate in schools. The Literacy in Future Technologies Artificial Intelligence, or LIFT AI Act, would empower the new director of the National Science Foundation (NSF) to make grant awards "on a merit-reviewed, competitive basis to institutions of higher education or nonprofit organizations (or a consortium thereof) to support research activities to develop educational curricula, instructional material, teacher professional development, and evaluation methods for AI literacy at the K-12 level," the bill says.
It defines AI literacy as using AI; specifically, "having the age-appropriate knowledge and ability to use artificial intelligence effectively, to critically interpret outputs, to solve problems in an AI-enabled world, and to mitigate potential risks." The bill is endorsed by the American Federation of Teachers, Google, OpenAI, Information Technology Industry Council, Software & Information Industry Association, Microsoft, and HP Inc. [...] The grant would support "AI literacy evaluation tools and resources for educators assessing proficiency in AI literacy," according to the bill. It would also fund "professional development courses and experiences in AI literacy," and the development of "hands-on learning tools to assist in developing and improving AI literacy." Most importantly for real-world implications, it would fund changing the existing curriculum "to incorporate AI literacy where appropriate, including responsible use of AI in learning."
Connexion
