Actualités — Page44

A class-action suit filed by an anonymous user, identified as John Doe, accuses Perplexity, the fast‑growing AI search platform, of breaching privacy promises. The complaint alleges that the company’s incognito feature fails to shield user conversations, instead funneling chat transcripts, IP addresses, email identifiers and location data to advertising partners such as Google and Meta. If the allegations prove true, the case could force tighter transparency standards across AI‑driven services.Lire la suite

Researchers analyzing 1,372 participants across more than 9,500 decision‑making trials discovered that people accepted AI‑generated answers that were wrong 73.2% of the time, while only overturning them in 19.7% of cases. The study links high trust in artificial‑intelligence systems to a greater likelihood of being misled, whereas individuals with higher fluid intelligence were more prone to question the AI. Authors warn that while reliance on AI can be advantageous when the technology is superior, the current tendency to treat AI output as authoritative creates a structural vulnerability in human judgment.Lire la suite

OpenAI announced a major executive reshuffle on April 3, 2026. COO Brad Lightcap will leave his operational role to head a new "special projects" unit reporting directly to CEO Sam Altman. CEO Fidji Simo disclosed she is taking several weeks of medical leave for a neuroimmune condition, while chief marketing officer Kate Rouch steps down to focus on cancer treatment. Denise Dresser, former Slack chief executive, assumes the chief revenue officer post, and co‑founder Greg Brockman will temporarily oversee product. The changes aim to preserve momentum on the company’s research and growth agenda.Lire la suite

Anthropic completed a $400 million stock purchase of Coefficient Bio, a stealth AI biotech firm founded by former Genentech researchers Samuel Stanton and Nathan C. Frey. The acquisition adds a ten‑person team focused on accelerating drug discovery to Anthropic’s health and life‑science division, following the company’s October launch of Claude for Life Sciences, an AI tool aimed at scientific research.Lire la suite

Anthropic has filed paperwork to create AnthroPAC, a political action committee funded by voluntary employee contributions up to $5,000 each. The PAC will target both parties in the upcoming midterm elections, donating to incumbent lawmakers and emerging candidates. The move signals the AI firm’s deeper foray into Washington’s lobbying arena amid an ongoing legal dispute with the Defense Department over the use of its models. Anthropic joins other tech companies that have collectively poured millions into election cycles in recent months.Lire la suite

OpenClaw, the AI‑driven automation tool that has amassed over 347,000 GitHub stars since its November debut, received emergency patches this week for three high‑severity bugs. The most dangerous, CVE‑2026‑33579, scores between 8.1 and 9.8 out of 10 and lets a low‑level pairing credential silently elevate to full administrative rights, giving a malicious actor unrestricted access to the host’s files, accounts and connected services.Lire la suite

OpenAI announced a series of senior‑level changes on Thursday as chief operating officer Brad Cox moves to a special‑projects role, former Slack CEO Denise Wasserman assumes his responsibilities, and chief marketing officer Kate Miller steps down to focus on cancer treatment. The updates come alongside CEO Sam Altman's disclosure that he will be on medical leave for several weeks due to a relapse of a neuroimmune condition.Lire la suite

Six months after renegotiating its contract with OpenAI, Microsoft unveiled MAI-Transcribe-1, MAI-Voice-1 and MAI-Image-2 on its Foundry platform. The new models, built by a ten‑person team, boast lower error rates, faster speeds and competitive pricing, giving the tech giant a functional AI stack independent of its former partner. The rollout underscores Microsoft’s new freedom to pursue "humanist superintelligence" and could reshape enterprise AI spending.Lire la suite

President Donald Trump’s effort to fast‑track AI data center construction is hitting a wall. Developers say aggressive tariffs on Chinese imports have crippled the supply chain for essential power‑equipment components, pushing delivery times from months to years. With U.S. manufacturers unable to meet demand, nearly half of the year’s planned facilities face delay or cancellation, jeopardizing the administration’s goal of keeping America ahead in the artificial‑intelligence race.Lire la suite

Google introduced the Gemma 4 family of open-weight AI models, offering four variants optimized for local execution and mobile devices. The two larger models—26B Mixture of Experts and 31B Dense—run unquantized on a single 80GB Nvidia H100 GPU and can be quantized for consumer GPUs. Smaller Effective 2B and Effective 4B models target smartphones and edge hardware, benefitting from collaboration with Qualcomm and MediaTek. Google also replaced its custom Gemma license with the Apache 2.0 license, giving developers greater freedom. The company claims Gemma 4 models are the most capable locally runnable AI systems, positioning them near the top of open AI model rankings.Lire la suite

Anthropic issued a DMCA notice to GitHub to remove a repository that contained leaked Claude Code client source code. The notice also listed nearly one hundred forks of that repository. GitHub’s automated processing interpreted the request as covering a broader network of about 8,100 similar forks, many of which were legitimate copies of Anthropic’s official public Claude Code repository. The over‑broad takedowns sparked backlash from developers, prompting Anthropic to ask GitHub to limit the removals to the specifically named URLs and to restore the other repositories.Lire la suite

ElevenLabs has quietly introduced ElevenMusic, a free iOS app that lets users create and discover AI‑generated songs using natural‑language prompts. The app offers daily limits, customizable song length and style, and a suite of listening features such as live stations and mood‑based mixes. A Pro tier priced at $9.99 per month expands limits to 500 tracks monthly and provides over 500 GB of storage. The launch reflects ElevenLabs' broader push beyond voice models into creative AI tools, positioning the company to compete with other AI music platforms.Lire la suite

OpenAI announced the purchase of TBPN, a daily livestream talk show that focuses on artificial intelligence and tech leadership. The acquisition aims to strengthen OpenAI's corporate communications and create a dedicated space for constructive conversations about AI. TBPN will retain editorial independence while operating under OpenAI's Strategy organization. The move comes amid heightened scrutiny of OpenAI's public image, large funding rounds, and ongoing debates over its defense contracts and revenue generation strategies.Lire la suite

Google Workspace’s AI‑powered video creator, Vids, has received a major upgrade that adds directable and custom avatars, a Veo 3.1 integration for on‑demand clip generation, direct YouTube export, and a built‑in screen‑recording Chrome extension. These tools let users type commands to control avatars, generate short video snippets from prompts, and streamline the production workflow, reducing the need for multiple apps and manual editing.Lire la suite

Google has added several new capabilities to its Vids video‑editing app. Users can now direct AI avatars with natural‑language prompts, customize appearance, apparel and backgrounds, and have avatars interact with props. The update also brings the Veo 3.1 video‑generation model, which creates eight‑second clips, and allows direct export of finished videos to YouTube as private files. A new Chrome screen‑recording extension lets users capture screen audio or video. Google offers 10 free generations per month to all users, while AI Ultra and Workspace AI Ultra accounts can generate up to 1,000 Veo videos each month.Lire la suite

Microsoft AI, the research arm of the tech giant, announced the rollout of three foundational multimodal models—MAI-Transcribe-1, MAI-Voice-1, and MAI-Image-2. The transcription model supports 25 languages and is 2.5 times faster than Azure Fast. The voice model can generate a minute of audio in one second and allows custom voice creation. The image model, originally unveiled on MAI Playground, expands Microsoft’s AI portfolio and is priced to be cheaper than competing offerings from Google and OpenAI. The launch underscores Microsoft’s commitment to building its own AI stack while maintaining its partnership with OpenAI.Lire la suite

A team led by KAIST has demonstrated that artificial‑intelligence models can be reverse‑engineered by capturing faint electromagnetic emissions from GPUs during normal operation. Using a small antenna hidden in a bag, the researchers collected traces from as far as six meters away, even through walls, and reconstructed key architectural details of AI systems with high accuracy. The technique, called ModelSpy, highlights a new physical‑layer vulnerability that bypasses traditional software and network defenses, raising concerns for companies that consider AI model designs as core intellectual property.Lire la suite

Microsoft announced three new artificial‑intelligence models: a voice model that can generate up to 60‑second audio clips, a transcription model that converts recordings into text in 25 languages, and a second‑generation image model that delivers faster, more realistic results. The models are now available in Microsoft’s Foundry and MAI playground, with plans to integrate the image model into Bing and PowerPoint. The rollout reflects Microsoft’s push to broaden its AI portfolio beyond text‑focused tools, complementing its Copilot suite and underscoring the company’s deep resources for enterprise‑grade generative media.Lire la suite

Meta is reevaluating its artificial‑intelligence roadmap as the proprietary ‘Avocado’ model, slated for a 2026 release, encounters performance setbacks and timeline pushes. The company, once champion of open‑source Llama models, appears poised to move toward closed‑source solutions after internal tests showed ‘Avocado’ lagging behind rivals such as Google’s Gemini series. Discussions about temporarily licensing Gemini have surfaced, highlighting a potential reliance on external technology. These developments raise questions about Meta’s long‑term AI direction and its ability to stay competitive in a rapidly evolving market.Lire la suite

A user turned to ChatGPT for advice on feeling less exhausted after a demanding week. The AI suggested three simple strategies: meal planning to eliminate daily decision‑making, establishing a consistent bedtime routine, and visualizing a future trip to create a sense of forward momentum. By following these suggestions for a week, the user reported reduced background stress, more alert mornings, and an overall lighter feeling, even though the advice was not groundbreaking. The experience highlights how clear, actionable guidance from AI can support everyday wellbeing.Lire la suite