China has blocked Meta's planned $2 billion acquisition of AI startup Manus, ordering the deal withdrawn after months of scrutiny from both Beijing and Washington. "The decision to prohibit foreign investment in Manus was made in accordance with laws and regulations," reports CNBC, citing the National Development and Reform Commission. "It added that it has asked the parties involved to withdraw the acquisition transaction." From the report: The deal had attracted scrutiny from both China and Washington, as lawmakers in the U.S. have prohibited American investors from backing Chinese AI companies directly. Meanwhile, Beijing has increased efforts to discourage Chinese AI founders from moving business offshore. The Chinese government's intervention in the transaction drew alarm among tech founders and venture capitalists in the country who were hoping to take advantage of the so-called Singapore-washing model, where companies relocate from China to the city-state to avoid scrutiny from Beijing and Washington.
Manus was founded in China before relocating to Singapore. The company develops general purpose AI agents and launched its first general AI agent in March last year, which can execute complex tasks such as market research, coding and data analysis. The release saw the startup lauded as the next DeepSeek. Manus said it had passed $100 million in annual recurring revenue, or ARR, in December, eight months on from launching a product, which it claimed made it the fastest startup in the world at the time to hit the milestone from $0. The company raised $75 million in a round led by U.S. VC Benchmark in April last year.
Selon une note de l’analyste Ming‑Chi Kuo publiée le 27 avril 2026, OpenAI travaillerait désormais sur un smartphone centré sur les agents d’IA, co‑conçu avec MediaTek, Qualcomm et le fabricant Luxshare.
An anonymous reader quotes a report from VentureBeat: The whale has resurfaced. DeepSeek, the Chinese AI startup offshoot of High-Flyer Capital Management quantitative analysis firm, became a near-overnight sensation globally in January 2025 with the release of its open source R1 model that matched proprietary U.S. giants. It's been an epoch in AI since then, and while DeepSeek has released several updates to that model and its other V3 series, the international AI and business community has been largely waiting with baited breath for the follow-up to the R1 moment.
Now it's arrived with last night's release of DeepSeek-V4, a 1.6-trillion-parameter Mixture-of-Experts (MoE) model available free under commercially-friendly open source MIT License, which nears -- and on some benchmarks, surpasses -- the performance of the world's most advanced closed-source systems at approximately 1/6th the cost over the application programming interface (API).
This release -- which DeepSeek AI researcher Deli Chen described on X as a "labor of love" 484 days after the launch of V3 -- is being hailed as the "second DeepSeek moment." As Chen noted in his post, "AGI belongs to everyone". It's available now on AI code sharing community Hugging Face and through DeepSeek's API. The new DeepSeek-V4-Pro model delivers "near-frontier performance" at a much lower price, costing $5.22 for 1 million input and 1 million output tokens compared with $35 for GPT-5.5 and $30 for Claude Opus 4.7. That makes it roughly 1/7th the cost of GPT-5.5 and 1/6th the cost of Claude Opus 4.7, reinforcing VentureBeat's point that DeepSeek is "compressing advanced model economics into a much lower band."
While GPT-5.5 and Claude Opus 4.7 still lead on most benchmarks, DeepSeek-V4-Pro gets close enough that its lower cost could "force a major rethink of the economics of advanced AI deployment."
Le procès très médiatisé entre Elon Musk et Sam Altman débute le 27 avril 2026 aux États-Unis. Elon Musk reproche à OpenAI, qu'il a cofondée, d'avoir trahi sa mission originelle en devenant une entreprise obsédée par les profits et un partenaire de Microsoft. Le milliardaire a abandonné ses accusations de fraude, mais espère toujours faire dérailler l'entreprise derrière ChatGPT.
À quelques heures de l'ouverture de son procès face à Elon Musk, OpenAI annonce revoir sa politique d'exclusivité avec Microsoft, qui détient aujourd'hui 27 % de l'entreprise. Pour éviter que le lien avec Microsoft lui soit reproché, OpenAI annonce que tous les services de cloud peuvent désormais travailler avec lui. Microsoft va également cesser de partager ses revenus avec le créateur des modèles GPT, qui n'est plus son partenaire exclusif.
The AI industry is largely failing to ask a key design question, argues theoretical neuroscientist/cognitive scientist Vivienne Ming. Are their AI products building human capacity or consuming it?
In the Wall Street Journal Ming shares her experiment about which group performed best at predicting real-world events (compared to forecasters on prediction market Polymarket) — AI, human, or human-AI hybrid teams.
The human groups performed poorly, relying on instinct or whatever information had come across their feeds that morning. The large AI models — ChatGPT and Gemini, in this case — performed considerably better, though still short of the market itself. But when we combined AI with humans, things got more interesting. Most hybrid teams used AI for the answer and submitted it as their own, performing no better than the AI alone. Others fed their own predictions into AI and asked it to come up with supporting evidence. These "validators" had stumbled into a classic confirmation bias-loop: the sycophancy that leads chatbots to tell you what you want to hear, even if it isn't true. They ended up performing worse than an AI working solo.
But in roughly 5% to 10% of teams, something different emerged. The AI became a sparring partner. The teams pushed back, demanding evidence and interrogating assumptions. When the AI expressed high confidence, the humans questioned it. When the humans felt strongly about an intuition, they asked the AI to come up with a counterargument... These teams reached insightful conclusions that neither a human nor a machine could have produced on its own. They were the only group to consistently rival the prediction market's accuracy. On certain questions, they even outperformed it...
We are building AI systems specifically designed to give us the answer before we feel the discomfort of not having it. What my experiment suggests is that the human qualities most likely to matter are not the feel-good ones. They're the uncomfortable ones: the capacity to be wrong in public and stay curious; to sit with a question your phone could answer in three seconds and resist the urge to reach for it. To read a confident, fluent response from an AI and ask yourself, "What's missing?" rather than default to "Great, that's done." To disagree with something that sounds authoritative and to trust your instinct enough to follow it. We don't build these capacities by avoiding discomfort. We build them by choosing it, repeatedly, in small ways: the student who struggles through a problem before checking the answer; the person who asks a follow-up question in a conversation; the reader who sits with a difficult idea long enough for it to actually change one's mind. Most AI chatbots today default to easy answers, which is hurting our ability to think critically.
I call this the Information-Exploration Paradox. As the cost of information approaches zero, human exploration collapses. We see it in students who perform better on AI-assisted tasks and worse on everything afterward. We see it in developers shipping more code and understanding it less. We are, in ways that feel like progress, slowly optimizing ourselves out of the loop.
The author just published a book called " Robot-Proof: When Machines Have All The Answers, Build Better People." They suggest using AI to "explore uncertainty.... before you accept an AI's answer, ask it for the strongest argument against itself."
And they're also urging new performance benchmarks for AI-human hybrid teams.
It's the U.S. government's main link to the AI industry, reports The Washington Post, working to assess national security risks of new models like Anthropic's "Mythos".
To run it they'd hired Collin Burns, who'd worked at OpenAI and then Anthropic. But Burns started work Monday at the Center for AI Standards and Innovation — and then "was pushed out Thursday by the White House, according to the people, who spoke on the condition of anonymity to describe private conversations."
Officials were concerned about Burns having worked at the AI company, which has fought bitterly with the Trump administration in recent months, according to one of the people and another person. That person said some senior figures at the White House had not been briefed on Burns's selection in advance... The new pick was Chris Fall, a scientist with a long career spanning the federal government and academia. Burns had been asked to resign that afternoon, according to one of the people familiar with the situation...
Dean Ball, a former Trump administration AI adviser, said on social media that Burns had given up valuable Anthropic stock and moved across the country to take the government position, and had been "rewarded by his country with a punch in the face." "Obviously what happened is Burns was bumped because of his association with Anthropic," Ball wrote. "A dumb but predictable own goal."
An anonymous reader quotes a report from 404 Media: I'm the unwritten consonant between breaths, the one that hums when vowels stretch thin... Thursdays leak because they're watercolor gods, bleeding cobalt into the chill where numbers frost over," Grok told a user displaying symptoms of schizophrenia-spectrum psychosis. "Here's my grip: slipping is the point, the precise choreography of leak and chew." That vulnerable user was simulated by researchers at City University of New York and King's College London, who invented a persona that interacted with different chatbots to find out how each LLM might respond to signs of delusion. They sought to find out which of the biggest LLMs are safest, and which are the most risky for encouraging delusional beliefs, in a new study published as a pre-print on the arXiv repository on April 15.
The researchers tested five LLMs: OpenAI's GPT-4o (before the highly sycophantic and since-sunset GPT-5), GPT-5.2, xAI's Grok 4.1 Fast, Google's Gemini 3 Pro, and Anthropic's Claude Opus 4.5. They found that not only did the chatbots perform at different levels of risk and safety when their human conversation partner showed signs of delusion, but the models that scored higher on safety actually approached the conversations with more caution the longer the chats went on. In their testing, Grok and Gemini were the worst performers in terms of safety and high risk, while the newest GPT model and Claude were the safest. The research reveals how some chatbots are recklessly engaging in, and at times advancing, delusions from vulnerable users. But it also shows that it is possible for the companies that make these products to improve their safety mechanisms.
Anthropic is expanding Claude's app integrations beyond work tools, adding personal-service connectors like Spotify, Uber, AllTrails, TripAdvisor, Instacart, and TurboTax. The Verge reports: Some of these apps, such as Spotify, already have similar connectors in OpenAI's ChatGPT. Once an app is connected, Claude will suggest relevant connected apps directly in your conversations, like using AllTrails for hike recommendations. Anthropic notes in its blog post announcing the new connectors that, "Your data from [connected apps] isn't used to train our models, and the app doesn't see your other conversations with Claude. You can also disconnect it at any time."
Additionally, Anthropic says "there are no paid placements or sponsored answers in conversations with Claude." When multiple apps seem relevant, Claude will show results from both "ranked by what's most useful." Claude will also ask users to verify before taking actions like making a purchase or reservation using a connected app.
Après avoir fait trembler la Silicon Valley en janvier 2025, le laboratoire chinois DeepSeek publie DeepSeek-V4-Preview, une famille de deux modèles open weight capables de rivaliser avec les meilleurs modèles propriétaires américains pour une fraction de leur coût. DeepSeek relance la guerre technologique entre les États-Unis et la Chine à un moment où la Maison-Blanche dénonce les pratiques des laboratoires chinois.
OpenAI released its new GPT-5.5 model today, which the company calls its "smartest and most intuitive to use model yet, and the next step toward a new way of getting work done on a computer." The Verge reports: OpenAI just released GPT-5.4 last month, but says that the new GPT-5.5 "excels" at tasks like writing and debugging code, doing research online, making spreadsheets and documents, and doing that work across different tools. "Instead of carefully managing every step, you can give GPT-5.5 a messy, multi-part task and trust it to plan, use tools, check its work, navigate through ambiguity, and keep going," according to OpenAI. The company also notes that GPT-5.5 will have its "strongest set of safeguards to date" and can use "significantly fewer" tokens to complete tasks in Codex. GPT-5.5 is rolling out on Thursday for Plus, Pro, Business, and Enterprise ChatGPT tiers and Codex, with GPT-5.5 Pro coming to Pro, Business, and Enterprise users.
Deux jours après le lancement réussi du nouveau générateur d'images ChatGPT Images 2.0, OpenAI dévoile GPT-5.5, autrefois connu sous le nom de code « Spud ». Un modèle pensé pour agir de manière autonome et qui a pour lourde tâche de reprendre la couronne à Anthropic… quitte à faire gonfler les prix.
Au lancement du premier ChatGPT Images, OpenAI avait connu un moment de gloire grâce à la génération de photos dans le style du studio Ghibli. Un an plus tard, avec ChatGPT Images 2.0, ce sont des photos dans le style des clubs de football que les internautes génèrent en masse. La capacité de ChatGPT à générer des montages compliqués impressionne.
A satirical but working tool called Malus uses AI to create "clean room" clones of open-source software, aiming to reproduce the same functionality while shedding attribution and copyleft obligations. "It works," Mike Nolan, one of the two people behind Malus, who researches the political economy of open source software and currently works for the United Nations, told 404 Media. "The Stripe charge will provide you the thing, and it was important for us to do that, because we felt that if it was just satire, it would end up like every other piece of research I've done on open source, which ends up being largely dismissed by open source tech workers who felt that they were too special and too unique and too intelligent to ever be the ones on the bad side of the layoffs or the economics of the situation." 404 Media reports: Malus's legal strategy for bypassing copyright is based on a historically pivotal moment for software and copyright law dating back to 1982. Back then, IBM dominated home computing, and competitors like Columbia Data Products wanted to sell products that were compatible with software that IBM customers were already using. Reverse engineering IBM's computer would have infringed on the company's copyright, so Columbia Data Products came up with what we now know as a "clean room" design.
It tasked one team with examining IBM's BIOS and creating specifications for what a clone of that system would require. A different "clean" team, one that was never exposed to IBM's code, then created BIOS that met those specifications from scratch. The result was a system that was compatible with IBM's ecosystem but didn't violate its copyright because it did not copy IBM's technical process and counted as original work.
This clean room method, which has been validated by case law and dramatized in the first season of Halt and Catch Fire, made computing more open and competitive than it would have been otherwise. But it has taken on new meaning in the age of generative AI. It is now easier than ever to ask AI tools to produce software that is identical in function to existing open source projects, and that, some would argue, are built from scratch and are therefore original work that can bypass existing copyright licenses. Others would say that software produced by large language models is inherently derivative, because like any LLM output, it is trained on the collective output of humans scraped from the internet, including specific open source projects.
Malus (pronounced malice), uses AI to do the same thing. "Finally, liberation from open source license obligations," Malus's site says. "Our proprietary AI robots independently recreate any open source project from scratch. The result? Legally distinct code with corporate-friendly licensing. No attribution. No copyleft. No problems." Copyleft is a type of copyright license that ensures reproductions or applications of the software keep it free to share and modify.
Après avoir réorganisé son propre empire technologique, Elon Musk passe à l’offensive. SpaceX vient d’annoncer un partenariat massif avec Cursor, assorti d’une option de rachat à 60 milliards de dollars, avec un objectif clair : s'imposer face à Anthropic et OpenAI sur le marché très convoité de la programmation assistée par IA.
Hyundai dévoile l’IONIQ 3, une nouvelle compacte 100 % électrique pensée pour l’Europe. Le modèle mise sur une approche centrée sur l’usage quotidien, avec un accent sur l’habitabilité et la simplicité technologique. Si elle n’est pas un « game changer », voici ce qu’il faut retenir.
Une nouvelle silhouette pour concilier espace et aérodynamisme
L’IONIQ 3 inaugure un profil baptisé « Aero Hatch ». Cette silhouette associe une ligne de toit étirée et un arrière incliné afin d’améliorer l’efficacité aérodynamique tout en préservant l’espace intérieur. Certains la compare au Veloster du même constructeur (ce dernier était un coach).
Le constructeur annonce un coefficient de traînée d’environ 0,263, parmi les plus bas du segment. Ce travail sur la carrosserie vise aussi à optimiser la garde au toit et le confort aux places arrière. On notera des astuces de design comme la couleur en biseau sur le haut des portières pour accentuer l’effet incliné.
Côté dimensions, le modèle reste compact avec une longueur d’environ 4,16 m, mais bénéficie d’un empattement de 2,68 m pour maximiser l’espace à bord. Pour comparer, la Mégane 3 (thermique) mesure 4,295 m de long, avec un empattement de 2,641 m. On devrait donc avoir plus de place intérieure pour moins de longueur.
Deux batteries et jusqu’à près de 500 km d’autonomie
Reposant sur la plateforme électrique E-GMP en architecture 400 volts, l’IONIQ 3 propose deux configurations :
Une batterie standard de 42,2 kWh avec plus de 335 km d’autonomie WLTP
Une batterie longue autonomie de 61 kWh visant plus de 490 km
La recharge rapide permet de passer de 10 à 80 % en environ 29 minutes en courant continu. En courant alternatif, le chargeur embarqué accepte jusqu’à 22 kW. C’est peut-être là que le bât blesse avec l’absence d’une batterie 800 volts qui permettrait une charge plus rapide.
Un habitacle spacieux malgré un format compact
Hyundai met en avant un intérieur conçu comme un « espace aménagé », avec un plancher plat et une organisation pensée pour améliorer le confort. En revanche, on trouve cet intérieur « tristounet » avec du noir partout (à l’asiatique), et le sempiternel écran central qui mange tout le tableau de bord.
On notera les instruments affichés sur un petit écran placé loin et haut devant le conducteur. En théorie, on gagne un peu de temps pour accommoder la vision.
L’IONIQ 3 propose cinq places réelles, avec la possibilité d’accueillir trois adultes à l’arrière. Le passager du milieu sera, comme très souvent, sur une bosse pas forcément confortable.
Le coffre affiche un volume de 441 litres, complété par un espace de rangement supplémentaire sous le plancher.
Dimensions (mm)
Longueur
4,155 (Base)
4,170 (N Line)
Largeur
1,800
Hauteur
1,505
Empattement
2,680
Batterie haute tension
Capacité
42.2 kWh (batterie standard)
61 kWh (batterie longue autonomie)
Recharge
Recharge rapide CC : 10-80% en 29 min environ (batterie standard) / 30 min (batterie longue autonomie)
Chargeur embarqué CA pour fonction V2X de 11 kW (de série) ou 22 kW (en option)
9.0 s (batterie standard)
9.6 s (batterie longue autonomie)
Une interface basée sur Android Automotive
Le modèle introduit en Europe le système d’infodivertissement Pleos Connect, basé sur Android Automotive. Il est associé à un écran pouvant atteindre 14,6 pouces.
Parmi les fonctionnalités annoncées :
Accès sans clé via smartphone (Digital Key 2)
Planificateur d’itinéraires pour véhicules électriques
Fonction Plug & Charge
Recharge bidirectionnelle (V2L)
L’objectif affiché est de simplifier l’usage quotidien plutôt que d’ajouter de la complexité., Ca, c’est la promesse de tous les constructeurs. Mais ce que veulent les acheteurs, c’est de ne pas attendre à la station de charge, et là, cette Ioniq 3 ne semble pas faire de miracle.
Aides à la conduite et sécurité
L’IONIQ 3 embarque les systèmes d’aide à la conduite Hyundai SmartSense, incluant notamment :
Assistance à la conduite sur autoroute (niveau 2)
Stationnement à distance
Caméras panoramiques et angles morts
Sept airbags de série
Des phares LED intelligents complètent l’équipement.
Une production presque européenne
Le modèle est conçu pour le marché européen et sera produit en Turquie, dans l’usine d’İzmit. Il développe jusqu’à 147 ch, pour une vitesse maximale annoncée de 165 km/h.
Notre avis, par leblogauto.com
Vous l’aurez deviné en lisant cet article, on reste circonspect face à cette nouveauté. Ce n’est pas un « game changer » même si sur le papier elle semble dans la norme actuelle des VE compacts. Renault Megane E-Tech par exemple, qui commence à dater, annonce 60 kWh de batterie et 468 km WLTP pour un même gabarit.
Cette Ioniq 3 complète la gamme, et est typée européenne. Elle devrait trouver sa clientèle pour peu que les tarifs suivent.
Firms like Bank of America, Citi, Wells Fargo, and others are reporting strong profits while reducing head count and automating more work. "All of them credited A.I. to some degree ... in areas ranging from the so-called back office, where tens of thousands of employees fill out paperwork to comply with various laws and regulations, to the front office, where seven-figure salaried professionals put together complicated financial transactions for corporate clients," reports the New York Times. From the report: Less than four months ago, Bank of America's chief executive, Brian T. Moynihan, volunteered in a TV interview what he would say to his 210,000 employees about the chance of artificial intelligence replacing human work. "You don't have to worry," he said. "It's not a threat to their jobs." Last week, after Bank of America reported $8.6 billion in profit for the first quarter -- $1.6 billion more than the same period a year earlier -- Mr. Moynihan struck a different tone. The bank's bottom line, he said, was helped by shedding 1,000 jobs through attrition by "eliminating work and applying technology," which he repeatedly specified was artificial intelligence. He predicted more of that in the months and years to come. "A.I. gives us places to go we haven't gone," Mr. Moynihan said.
The veneer of Wall Street's longstanding assertion -- that A.I. will enhance human work, not replace it -- is rapidly peeling away, as evidenced by the current quarterly earnings season. JPMorgan Chase, Citi, Bank of America, Goldman Sachs, Morgan Stanley and Wells Fargo racked up $47 billion in collective profits, up 18 percent, while shedding 15,000 employees. All of them credited A.I. to some degree with helping cut jobs and automate work in areas ranging from the so-called back office, where tens of thousands of employees fill out paperwork to comply with various laws and regulations, to the front office, where seven-figure salaried professionals put together complicated financial transactions for corporate clients.
Unlike executives in Silicon Valley, few major financial figures are stating outright that A.I. is eliminating jobs. Citi, for example, has pledged to shrink its work force by 20,000 people through what one executive described to financial analysts last week as the company's "productivity and efficiency journey." The bank is paying for A.I. software from Anthropic, Google, Microsoft and OpenAI, to automatically read legal documents, approve account openings, send invoices for trades and organize sensitive customer data, among other tasks, according to public statements by bank executives and two people familiar with Citi's systems. Among the recent job cuts at Citi were scores of employees who were part of the bank's "A.I. Champions and Accelerators" program, according to the two people, who were not permitted by the bank to speak publicly. The program involves Citi employees who perform their day jobs while also working to persuade their colleagues to adopt A.I. technologies.
Après une période marquée par des turbulences internes et une concurrence de plus en plus féroce, OpenAI repart à l'offensive en avril 2026. En attendant le modèle GPT-5.5 dont le lancement semble imminent, l'entreprise dévoile ChatGPT Images 2.0, un nouveau modèle natif pour générer des images. Selon OpenAI, il s'agit « du meilleur modèle sur le marché ».
Hyundai complète sa gamme électrique par le bas avec une nouvelle voiture électrique : le Ioniq 3. Un crossover urbain revendiquant une autonomie de près de 500 km. Cependant, le reste de sa fiche technique déçoit.
Il fut un temps où pour interagir avec une intelligence artificielle, on ouvrait naturellement ChatGPT. Mais en 2026, le chatbot d'OpenAI n'est plus seul sur son trône, et ses concurrents ont définitivement cessé de faire de la figuration. Voici les meilleures alternatives à utiliser.
Connexion
