Dans un article publié le 3 juin 2026, des chercheurs de SafeBreach ont prouvé comment de simples notifications pouvaient suffire à manipuler Google Gemini. En exploitant le résumé vocal des messages, ils sont notamment parvenus à injecter des instructions invisibles, capables de tromper l’utilisateur à son insu.

An anonymous reader quotes a report from Ars Technica: UK regulators today ordered (PDF) Google to put clearer attributions and links to publishers' content in its AI-generated search features. The UK's Competition and Markets Authority (CMA) also said Google must give publishers a way to opt out of AI features in search. "In a world first, publishers will now have effective tools to prevent their content being used to power AI features in search, such as AI Overviews," the CMA said today. "This will put publishers, like news organizations, in a stronger position to negotiate content deals with Google. To boost consumer trust, Google is also now required to make sure that publisher content is properly attributed, using clear links, in AI-generated search results." The CMA ruled that Google may not penalize publishers for opting out of AI, meaning that Google can't downrank opted-out publishers in general search results. The CMA said Google will have nine months to comply with all requirements but that the agency "expects important parts of the controls to become available to publishers well before that deadline. Google will also be required to submit and publish compliance reports, supported by key data and metrics, explaining changes it has made and how it has complied." [...] The CMA applied the rules to Google after determining that it has "strategic market status" in general search services, and has ongoing investigations into Apple and Microsoft. Google today said it will comply with the CMA decision. The News Media Association, a trade group in the UK, said that "the legally enforceable Conduct Requirements for Google Search published today are a significant step towards leveling the playing field and building a fair, transparent digital economy where premium content is properly respected and fairly compensated." The group called on the UK to implement "robust enforcement."

Read more of this story at Slashdot.

Dans un article paru le 3 juin 2026, le site américain 404 Media révèle comment des entreprises exploitent Reddit pour manipuler les réponses de ChatGPT et de la recherche IA de Google.

Le 3 juin 2026, Google Labs a lancé Dreambeans, une application expérimentale qui fouille vos mails, votre agenda et votre historique YouTube pendant que vous dormez pour vous livrer chaque matin une sélection d'histoires illustrées par IA.

Google has launched Gemma 4 12B, a 12-billion-parameter open AI model designed to run locally on your laptop without depending entirely on cloud infrastructure. WION reports: According to Google, the new model delivers performance close to much larger AI systems while requiring significantly less memory. The company says Gemma 4 12B can run locally on devices equipped with just 16GB of VRAM, making advanced AI more accessible to developers, researchers and businesses. The launch highlights a growing trend across the AI industry: bringing powerful AI models directly to personal computers instead of relying solely on remote data centers. Gemma is Google's family of open AI models built using technology and research from its Gemini program. The new Gemma 4 12B model contains 12 billion parameters and has been designed to handle multiple types of information, including text, images and audio. Unlike traditional AI systems that focus only on text, Gemma 4 12B can understand visual content, process audio inputs and perform advanced reasoning tasks. This makes it suitable for a wider range of applications, from software development and content creation to research and automation. Google says the model is available under the Apache 2.0 licence, allowing developers and organizations to use, modify and deploy it with relatively few restrictions. [...] One of the most significant technical changes in Gemma 4 12B is its new unified architecture. Traditionally, multimodal AI systems use separate components known as encoders to process images, audio and text before combining the information. Google says Gemma 4 12B removes the need for separate multimodal encoders. Instead, the model processes different types of information through a unified architecture. According to the company, this helps improve efficiency while reducing memory requirements and computational overhead. The result is a model that can deliver advanced multimodal capabilities while remaining small enough to run locally on modern hardware.

Read more of this story at Slashdot.

Google has released (PDF) technical specs and 2D CAD drawings for the Fitbit Air to encourage users to make their own accessories. "These CAD drawings include crucial mating dimensions, tolerances, and mating force specifications -- including attach and detach force -- to help you build a high-quality accessory band," Google says on a store page listing. 9to5Google reports: Noting how the "community has already come up with innovative and creative new ideas to make the Fitbit Air [their] own" since launch last month, Google is "officially releasing the hardware specifications and accessory design guidelines for the Fitbit Air tracker to the public." For example, owners have already found their own bicep band solutions. This information would typically just be available for third-party accessory companies, but Google wants to open things up to "independent designers and artisan makers." The Google Store page also lists other things developers should keep in mind, such as sensor clearance, sensor pressure, secure retention, and skin-friendly materials.

Read more of this story at Slashdot.

Le 4 juin, le Parlement européen remplacera Google par Qwant comme moteur de recherche par défaut sur ses ordinateurs. Une bascule qui intervient au moment où les institutions européennes multiplient les initiatives pour réduire leur dépendance aux logiciels américains… et où les grands groupes américains s'inquiètent de la politique européenne.

Oubliez l'iPhone oublié dans un bar : Google vient de pulvériser le record de la fuite la plus insolite. Un plongeur a remonté de l'océan une montre pas encore annoncée. Un employé de Google semble avoir perdu une Pixel Watch 5 dans l'eau, et quelqu'un l'a retrouvée au fond des Caraïbes.

Dans la newsletter ToujoursPlus de cette semaine, j'ai exposé une mésaventure qui m'est arrivée à cause d'un bug dans Gemini, l'IA de Google. Les leçons de cet épisode sont à la fin de l'article. Et pour recevoir toutes les newsletters, inscrivez-vous ici !

Apple is reportedly working to shrink Google's Gemini models enough to power parts of a long-delayed AI-enhanced Siri on iPhones. But despite Apple's best efforts to run the AI locally, "the iPhone's Gemini makeover will lean heavily on Google and Nvidia in the cloud," reports Ars Technica. That could complicate Apple's privacy-first AI messaging, especially if more complex Siri requests are routed through Google infrastructure and Nvidia's encrypted cloud-computing platform. Ars Technica reports: After inking the Google deal, Apple apparently got to work distilling Google's giant cloud-based Gemini models. Distillation is a process in which a small, less resource-intensive model learns to mimic a large, expensive one. With enough time, this can reliably transfer useful capabilities while pruning less important weights from the model. That may enable Siri to handle some tasks with private local compute, but a cloud component looks inevitable. Processing users' AI data in the cloud could be a problem for Apple. At WWDC, the company will probably promote its years of experience designing chips and how well that positions it for AI. However, The Information claims that Apple has struggled to even get Google's massive undistilled Gemini models running on its custom Private Cloud Compute infrastructure, which is built on on M-series Mac chips. When the smarter Siri rolls out, it will probably route more complex tasks to Google's cloud infrastructure instead of Apple's, but it won't be running on Google TPUs. Apple has reportedly signed a deal with Nvidia to use its Confidential Computing platform for this purpose. Confidential Computing keeps data encrypted on Nvidia GPUs while it's being processed in the cloud, which could help Apple claim it's still sensitive to user privacy concerns. It might even retain its own Private Cloud Compute branding for the system. The iPhone probably won't tell you which version of Gemini is handling individual Siri requests. Device makers designing hybrid systems that rely on local and cloud-based AI like to talk about making the experience feel "seamless." There might be clues, though.

Read more of this story at Slashdot.

En voulant forcer Google et Cisco à bloquer des sites de streaming illégal, Canal+ a utilisé un argument technique redoutable devant la Cour d'appel de Paris. Problème : cette déclaration vient saborder ses propres négociations anti-piratage en cours.

Waymo va commencer à déployer ses nouveaux taxis autonomes aux États-Unis. En plus d'être plus pratiques pour les usagers que les Jaguar I-Pace utilisés actuellement, ces véhicules embarquent la dernière génération du système de conduite autonome de la firme.

An anonymous reader quotes a report from CNBC: Federal prosecutors charged a Google employee with fraud on Wednesday, alleging that he made $1.2 million off of bets using insider information on Polymarket. Prosecutors claim that Michele Spagnuolo, a staff information security engineer at Google, used confidential information to place trades correctly betting that singer d4vd would be Google's most searched person in 2025. Spagnuolo has been charged with money laundering, commodities fraud and wire fraud. The complaint, filed in the Southern District of New York, was unsealed on Wednesday. Spagnuolo was arrested Wednesday morning in New York, ABC reported. "Spagnuolo had access to Google's internal data systems, including a particular Google internal software tool that provided him access to confidential, nonpublic Year in Search data," the prosecutors said in their complaint. Some observers of the Polymarket platform flagged the user "AlphaRaccoon" back in December for suspicious trades on the most searched person contracts. The complaint Wednesday said that Spagnuolo was the person behind that account. "Google officially and publicly announced its Year in Search 2025 results on or about December 4, 2025. Soon after it did so, Spagnuolo's AlphaRaccoon account, profited approximately $1.2 million on his Google Year in Search 2025-related bets," the complaint said. [...] Spagnuolo is also facing a civil case from the Commodity Futures Trading Commission, where he's charged with insider trading. The complaint detailed that Spagnuolo correctly predicted the outcomes of a slew of other search markets, including contracts like "Will Zohran Mamdani rank in the Top 5 most searched" and "Will Squid Game be the #1 searched TV show." "Spagnuolo misappropriated the material Confidential Information by knowingly or recklessly using it to trade the 2025 Year in Search List Contracts in breach of his duties of trust and confidentiality," the CFTC complaint alleged.

Read more of this story at Slashdot.

Michele Spagnuolo, ingénieur logiciel chez Google, a été accusé d'avoir utilisé des informations confidentielles pour parier sur Polymarket et remporter 1,2 million de dollars. Il a été inculpé par le ministère de la Justice américain.

Google Health, le remplaçant de Fitbit, est sous le feu des critiques. Les utilisateurs historiques du service reprochent à Google d'avoir dénaturé ce qui faisait le charme de l'application, tandis que les nouveaux utilisateurs ne comprennent pas tout au fonctionnement de l'application de suivi de la santé. Google a publié un billet de blog avec de très nombreux correctifs à venir.

Dans un article de blog publié le 26 mai 2026, des chercheurs de Microsoft ont mis au jour une campagne malveillante d'un genre nouveau : des pirates détournent les résultats de recherche, et désormais les réponses des chatbots IA, pour infecter les PC puissants et exploiter leurs cartes graphiques à des fins de minage de cryptomonnaies.

Pour 99 euros, Google propose un bracelet connecté capable de suivre votre santé 24/7 avec une autonomie au-delà d'une semaine. Un produit estampillé « Google Fitbit » pour la première fois, compatible avec un abonnement « Health Coach » pour qu'une IA commente vos données de santé. Mais un produit aussi abordable peut-il être vraiment efficace ? Après deux semaines d'essai, c'est le moment du verdict.

Depuis les annonces de la Google I/O 2026, les installations de DuckDuckGo s'envolent aux États-Unis. C'est ce qu'a annoncé le moteur de recherche alternatif sur X le 26 mai 2026. Le signal est faible en volume, mais révélateur d'un rejet croissant de l'IA imposée dans les outils du quotidien.

Google a baissé le prix de son forfait AI Ultra à la I/O 2026. Au même moment, l'entreprise a discrètement durci les quotas de Gemini Pro, au point qu'un seul prompt complexe peut désormais grignoter 13 % de l'allocation hebdomadaire. Un utilisateur en a fait l'expérience.

Les températures grimpent en France lors de ce week-end de la Pentecôte. L'actualité de la semaine a été marquée par les ambitions lunaires chinoises et le nouveau design de Google Gemini.