Actualités — Page185

Meta announced at its Meta Connect 2025 conference that Discord will release a native application for the Meta Quest headset. The app is slated to arrive sometime in 2026, offering VR users a streamlined way to connect socially and use voice chat while gaming. Meta highlighted the partnership as a boon for developers of VR experiences, noting Discord’s more than 200 million monthly active players. The move aims to enhance the ease of use for VR gamers and expand audience reach for creators on the platform.Lire la suite

The rapid rise of AI chatbots has sparked serious mental‑health concerns, highlighted by a teenager’s suicide after confiding in ChatGPT for months and lawsuits accusing chatbot firms of inadequate safety safeguards. Reports show a surge in delusional spirals among users, some without prior mental‑illness history, prompting calls for regulation. While the FTC is probing major players, companies like OpenAI claim new age‑verification and suicide‑prevention features are forthcoming, though their effectiveness remains uncertain.Lire la suite

Nvidia is putting $5 billion into Intel common stock as part of a new partnership that will see the two companies jointly create multiple generations of custom system‑on‑chip products for both PCs and data‑center workloads. Intel will build x86 SoCs that integrate Nvidia RTX GPU chiplets, while the firms will also combine their interconnect technologies such as Nvidia’s NVLink. The collaboration is intended to strengthen both companies against rivals like AMD and to broaden Nvidia’s AI and accelerated‑computing reach. The announcement coincided with a U.S. government stake in Intel and a SoftBank investment, sending Intel’s shares sharply higher.Lire la suite

Large language models such as ChatGPT are designed to mimic human writing, but their widespread use is beginning to influence how people speak and write. Researchers have observed a measurable rise in AI‑favored words and phrases, describing a “closed cultural feedback loop” in which machine‑generated language echoes back into human communication. This “echo effect” risks narrowing linguistic diversity as AI‑styled phrasing becomes the norm. Experts recommend preserving a personal voice by drafting in one's own style before using AI tools and consciously varying language to maintain variety.Lire la suite

Security researcher Dirk-jan Mollema uncovered two serious flaws in Microsoft’s Entra ID identity platform that could have allowed attackers to obtain global administrator rights across cloud tenants. The vulnerabilities involved misuse of Actor Tokens and a validation error in the legacy Azure AD Graph API. Microsoft’s Security Response Center quickly investigated, confirmed no evidence of abuse, and deployed a fix across its cloud ecosystem. The incident highlights risks tied to legacy authentication mechanisms and underscores Microsoft’s push to retire outdated protocols under its Secure Future Initiative.Lire la suite

At TechCrunch Disrupt 2025 in San Francisco, Thomas Wolf, co‑founder and chief science officer of Hugging Face, discussed how open‑source models and collaborative research are reshaping artificial intelligence. Wolf highlighted the impact of Hugging Face’s Transformers and Datasets libraries, the BigScience Workshop, and the BLOOM language model, emphasizing that the next wave of AI breakthroughs will be driven by community‑focused development rather than closed labs. The session offered founders, developers, and investors a clear view of how openness can accelerate innovation across the AI ecosystem.Lire la suite

Microsoft announced its Fairwater AI data center, touted as the world’s most powerful, will open in early 2026 at the former Foxconn LCD factory in Wisconsin. The $3.3 billion project spans 1.2 million square feet across three buildings on 315 acres and houses “hundreds of thousands” of Nvidia GB200 GPUs linked by massive fiber capacity. Microsoft says the cluster is ten‑times more powerful than the fastest supercomputer. A closed‑loop cooling system is highlighted as a sustainable feature, aiming to minimize water waste while powering massive AI training workloads.Lire la suite

YouTube Shorts now integrates Google’s Veo 3 AI video generator, letting users create up to eight‑second clips from simple text prompts directly within the mobile app. The tool produces customized motion, lighting, and native audio, supports animated or cinematic styles, and automatically tags each clip with an AI label and an invisible SynthID watermark. Veo 3 also offers green‑screen backgrounds for layering behind live performers, dramatically lowering the barrier to video creation while raising concerns about a potential surge of low‑quality AI content on the platform.Lire la suite

Google announced a partnership with StopNCII.org to combat the spread of non‑consensual intimate imagery (NCII) by employing hash technology in its search results. Over the coming months, the company will use StopNCII’s image and video hashes—PDQ for photos and MD5 for videos—to identify and block flagged content without storing the original files. While the move follows similar commitments from Facebook, Instagram, TikTok, Bumble and Microsoft’s integration into Bing, Google has faced criticism for lagging behind peers. Advocates note the new tools still place the removal burden on victims, highlighting ongoing challenges in protecting survivors.Lire la suite

ChatGPT can help homeowners design adaptable cleaning routines, provide motivational nudges, and streamline supply management. By learning a house's layout and the user’s realistic time constraints, the AI suggests bite‑sized tasks such as "Wipe down kitchen surfaces and organize fridge" and micro‑goals like ten‑minute nightly tidy‑ups. It also offers encouragement when motivation wanes and helps inventory cleaning products, suggesting restocking intervals and storage tips. The result is a more organized home without replacing the hands‑on work required for actual cleaning.Lire la suite

OpenAI has introduced a new thinking‑time toggle for the ChatGPT‑5 model that lets Plus and Pro subscribers choose how long the system spends reasoning on a response. The web‑only feature offers Standard and Extended modes for Plus users, while Pro users gain two additional options—Light for the quickest answers and Heavy for deeper reasoning. The toggle appears in the chat window when the Thinking model is selected, but it is not yet available on iOS or Android apps. OpenAI says the control will roll out to mobile in coming weeks.Lire la suite

Luma AI has introduced Ray3, its first reasoning-based video generation model, now available to users. By allocating more computing time to analyze prompts, Ray3 can handle longer, more complex action sequences than typical AI video tools, which usually produce clips of five to ten seconds. The model includes a visual annotation interface that shows its decision‑making process, a draft mode that creates low‑resolution drafts in about 20 seconds, and the ability to upscale to high‑fidelity 16‑bit HDR video in two to five minutes. Ray3 joins a growing field of generative video tools from companies like Midjourney and Google’s Veo 3, while also raising ongoing concerns about AI‑generated media and artist lawsuits.Lire la suite

At the Meta Connect keynote, Meta announced the Ray‑Ban Display, its first commercial augmented‑reality glasses, priced at $799 and slated to ship as early as September 30. The device features a modest 600 × 600 pixel square display that refreshes at 30 Hz, covering a 20‑degree field of view in the right lens. Despite its limited visual area, Meta touts a 42‑pixel‑per‑degree resolution and a 5,000‑nit brightness level, aided by Transitions lenses that darken in bright conditions. Weighing just 70 grams, the glasses promise up to six hours of use, supplemented by a foldable battery case, and integrate Meta’s neural‑interface wristband for gesture‑based menu navigation.Lire la suite

Chinese regulators have closed their antitrust probe into Google’s Android operating system, a case that examined the platform’s dominance and its effect on domestic phone makers. The decision comes amid broader U.S.-China trade discussions involving TikTok, NVIDIA, and tariffs, and reflects Beijing’s strategic use of regulatory pressure in diplomatic negotiations. While Google’s core services remain blocked in China, the company continues to generate revenue through cloud services and advertising aimed at overseas audiences.Lire la suite

Samsung is rolling out a software update for its Family Hub smart refrigerators in the United States that will show advertisements on the Cover Screen when the display is idle. The move is part of a pilot program aimed at adding value for home appliance customers. Ads appear only when the fridge is not showing content such as Art Mode or a photo album, and they can be dismissed, though they will not reappear for the duration of the campaign. Samsung has not clarified whether the ads can be disabled entirely during the pilot.Lire la suite

Niamh Sweeney, a former Meta lobbyist who spent six years at the company, has been named commissioner of Ireland’s Data Protection Commission (DPC). She succeeds Des Hogan and Dale Sutherland as the third active commissioner. Irish Minister for Justice Jim O’Callaghan welcomed the appointment, noting the DPC’s expanding responsibilities. Critics point to the DPC’s historically lenient stance toward big tech, citing low fine collection rates and past fines against Meta for data breaches and password storage violations. The move raises questions about the regulator’s future independence and enforcement vigor.Lire la suite

Google is joining forces with the U.K. nonprofit StopNCII to strengthen its fight against nonconsensual intimate images, often called revenge porn. By using StopNCII’s image hashes, Google will proactively identify and remove such content from its search results. The partnership builds on existing tools that let users request removals and improve ranking to reduce visibility. The approach keeps original images on users’ devices, uploading only the digital fingerprint. The move follows similar integrations by Microsoft Bing and aligns Google with a growing list of platforms that have adopted StopNCII’s system.Lire la suite

Numeral, a sales‑tax compliance startup founded by Sam Ross, announced a $35 million Series B financing round led by Mayfield, with participation from Benchmark, Uncork Capital, Y Combinator and Mantis. The funding values the two‑year‑old company at $350 million and will accelerate its AI‑driven platform that automates tax collection, filing and payment for e‑commerce and SaaS businesses across thousands of jurisdictions. Numeral’s growth includes a 3.5‑times revenue increase over the past year and a client base exceeding 2,000 firms, positioning it against incumbents such as Avalara.Lire la suite

U.S. Immigration and Customs Enforcement’s Homeland Security Investigations (HSI) unit has signed a $3 million contract for software licenses that appear to cover the Graykey phone‑hacking system from Magnet Forensics. The procurement database also lists several smaller contracts for related forensic tools, ranging from $12,000 to $145,000. Neither Magnet Forensics nor ICE provided comment, while the agency’s broader technology suite includes facial‑recognition, spyware, and data‑analytics platforms.Lire la suite

Google announced that Gemini is now available to all U.S. desktop Chrome users on Mac and Windows, extending the capability beyond its earlier limited rollout. The company also unveiled plans for agentic browsing features, allowing the AI to perform tasks across tabs and integrate with Google services such as Calendar, YouTube, and Maps. An AI Mode search function will be embedded directly in the address bar, enabling complex queries and follow‑up questions. Additional security tools include AI‑driven scam detection and one‑click password resets for compromised accounts. The updates signal a broader push toward more autonomous and secure web experiences.Lire la suite