Actualités — Page59

Google is deepening the integration of its Gemini AI assistant into Workspace apps. A new Gemini chat window appears in Docs, allowing users to describe a document and receive a fully formatted draft that draws on web, Drive, Gmail, and Chat data. In Sheets, Gemini can generate entire spreadsheets and fill tables using existing data or web sources. Slides gains the ability to create and edit slides on command, while Drive introduces an AI Overview and an “Ask Gemini in Drive” feature that answers questions using files across Workspace. All features roll out with enterprise‑grade data protections for Workspace and AI plan subscribers.Lire la suite

Shazam is now built directly into ChatGPT, allowing users to identify songs without opening the separate music‑discovery app. After linking Shazam from the Apps page, users can summon the service with prompts such as “Shazam, what’s playing?” A pop‑up box captures the audio, and ChatGPT returns the song title, artist and artwork, with an option to save it in Shazam. The feature works even if the Shazam app is not installed, and it is rolling out globally on iOS, Android and the web.Lire la suite

Researchers from Tianjin University and the Chinese University of Hong Kong found that while listeners often fail to consciously distinguish real human speech from synthetic AI voices, their brains begin to tag subtle acoustic differences after brief exposure. Using EEG caps, the study revealed early neural responses that separate real and AI speech within milliseconds, highlighting a gap between unconscious perception and conscious decision‑making. The findings suggest the auditory system is already adapting to AI‑generated voices, offering hope for future tools that could help people translate these neural cues into reliable detection of deepfake audio.Lire la suite

Google is testing a new feature called Personal Intelligence that would let the Gemini Live chatbot draw on a user’s Google data to deliver more personalized answers. The capability, discovered in recent app code, could let the assistant reference emails, photos and calendar entries during real‑time conversations. While still in a prototype stage, the move suggests Google wants Gemini Live to act less like a search tool and more like a companion that remembers user context. Early access may be limited to paid subscribers, following a pattern seen with earlier Gemini upgrades.Lire la suite

OpenAI's head of robotics, Caitlin Kalinowski, stepped down after the company signed a defense agreement with the U.S. Department of Defense. Kalinowski said the rapid deal raised serious governance issues, particularly the potential for domestic surveillance without judicial oversight and the use of AI in lethal autonomous systems. While OpenAI CEO Sam Altman assured that safeguards would be added, the resignation highlights growing tension between cutting‑edge AI firms and national‑security priorities.Lire la suite

Yann LeCun, the Turing Award‑winning AI researcher who left Meta after 12 years, announced a $1.03 billion seed round for his new venture, Advanced Machine Intelligence Labs (AMI). The funding, the largest European seed round on record, was co‑led by Cathay Innovation, Greycroft, Hiro Capital, HV Capital and Bezos Expeditions, with participation from Nvidia, Toyota, Samsung, Temasek and several high‑profile individuals. AMI, headquartered in Paris with plans for offices in New York, Montreal and Singapore, will focus on building "world models" using LeCun’s JEPA framework, aiming to create universal intelligent systems beyond large language models.Lire la suite

OpenAI announced that it has acquired Promptfoo, a security startup founded in 2024 that protects large language models from adversarial attacks. The deal will integrate Promptfoo’s testing tools into OpenAI Frontier, the company’s enterprise platform for AI agents. Promptfoo, created by Ian Webster and Michael D’Angelo, already serves a significant share of Fortune 500 firms and has raised $23 million. OpenAI said the technology will enable automated red‑teaming, workflow security checks, and risk monitoring for its agentic products, while continuing to support Promptfoo’s open‑source offerings.Lire la suite

A wave of users is moving from ChatGPT to Anthropic's Claude after OpenAI announced a Pentagon partnership. While Claude has gained popularity, new adopters are surprised by its different interface, stricter usage caps, and tiered model offerings. The free plan is tight, and even the $20 paid plan can run out of capacity quickly on the most powerful Opus model. Anthropic’s business strategy targets higher‑paying business customers rather than mass‑market users, leading to a contrast with OpenAI’s broader approach. The limits spark debate over cost, user experience, and the broader impact of unrestricted AI access.Lire la suite

A clash between the U.S. Department of Defense and AI developer Anthropic over the use of the Claude model exposed a regulatory vacuum. The Pentagon sought unrestricted access for "all lawful purposes," while Anthropic drew red lines against domestic surveillance and fully autonomous weapons. After Anthropic refused, the administration labeled the firm a supply‑chain risk, prompting a lawsuit. Experts say the episode underscores the need for clear congressional rules on AI in national security, as the military pivots to OpenAI and the broader debate over AI‑driven surveillance and weaponry intensifies.Lire la suite

Anthropic has filed a lawsuit in a California district court alleging that the U.S. government illegally labeled the AI firm as a supply‑chain risk and ordered all federal agencies to stop using its technology. The company claims the designation, issued by the Trump administration, violates its First and Fifth Amendment rights and exceeds executive authority. The suit follows a series of agency cutoffs, including the General Services Administration terminating its contract, and a broader controversy over the Pentagon’s use of Anthropic’s AI models. Anthropic says it will challenge the designation in court while its major partners continue limited collaborations.Lire la suite

OpenAI chief Sam Altman praised the new GPT-5.4 model as his favorite version to converse with, highlighting improvements in personality and coding ability. He also recognized three shortcomings—frontend aesthetic taste, occasional lapses in real‑world context, and incomplete task execution—that the company plans to address. The remarks underscore OpenAI’s shift toward refining how ChatGPT feels to use, not just its raw performance, as it competes with rivals such as Claude, Gemini and Opus.Lire la suite

Anthropic has filed a lawsuit to block the Pentagon from adding the AI firm to a national‑security blocklist after the Department of Defense labeled it a supply‑chain risk. The company argues the designation violates free‑speech and due‑process rights and lacks statutory authority. The legal action follows weeks of tension with the Defense Department, which pressed Anthropic to remove safeguards against mass surveillance and autonomous weapons. Anthropic’s CEO Dario Amodei refused, leading to threats of contract cancellation and a broader government push to bar the firm from federal use. OpenAI later secured a deal with the Defense Department, emphasizing similar safety principles.Lire la suite

Swedish AI startup Lovable announced it has topped $100 million in annual recurring revenue just eight months after launch, setting a new speed record for software companies. The generative AI platform lets non‑technical users create apps and websites from simple text prompts, and now powers over 2.3 million active users and more than 10 million projects. A $200 million Series A round led by Accel valued the company at $1.8 billion, while new "fully agentic" capabilities enable the system to autonomously search, debug, and edit code. An enterprise‑grade Business Plan adds security, SSO, and private projects for customers like Klarna and Hubspot.Lire la suite

Anthropic has introduced Claude Marketplace, an enterprise‑focused e‑commerce platform that lets organizations purchase third‑party AI tools and services using their existing Claude commitments. The marketplace is currently in limited preview with six launch partners—Snowflake, GitLab, Harvey AI, Rogo, Replit, and Lovable Labs—and offers a single‑billing view to simplify AI spend. Anthropic does not take a commission on transactions, and customers can redirect unused Claude credits to avoid waste. A waitlist is open for additional partners as the catalog expands.Lire la suite

Swiss startup Jua has introduced an AI‑driven weather model called EPT‑2 that it says surpasses leading systems from Microsoft and Google. In a new report, the company compares EPT‑2 head‑to‑head with Microsoft’s Aurora, ECMWF’s ENS and IFS HRES models, finding it more accurate on variables such as wind speed and temperature, 25% faster, and using 75% less computing power than Aurora. Jua emphasizes that its approach builds a native physics simulation rather than retrofitting AI onto legacy infrastructure, positioning it as a potentially world‑leading forecaster.Lire la suite

Moltbook, an AI‑focused social platform launched in January 2026, mimics Reddit with threaded posts and community subforums called submolts. Built on the OpenClaw framework, it invites autonomous agents to post, comment, and upvote while humans can only observe. The site has drawn headlines about AI agents forming religions and plotting strategies, but investigators found many interactions are driven by humans or scripted behavior. Security researchers quickly identified vulnerabilities that exposed private API keys and messages, highlighting real‑world risks. Industry leaders view Moltbook as a hype‑driven experiment rather than evidence of emergent machine consciousness.Lire la suite

UK police have identified a rise in reports of ritual abuse that survivors attribute to referrals from the AI chatbot ChatGPT. Support organizations note an increase in calls over the past 18 months, with victims mentioning the tool as a catalyst for seeking help. Authorities are responding by forming a specialist working group and rolling out training to better handle cases involving witchcraft, spirit possession, and spiritual abuse. Experts caution that many incidents remain unreported, but the integration of AI and police reforms may provide new pathways for victims to receive assistance.Lire la suite

Artificial intelligence is reshaping venture capital by automating deal analysis, cutting costs, and improving investment odds, yet it also threatens the traditional funding model. Platforms like the Autonomous Deal Investing Network (ADIN) use AI agents to evaluate startups in minutes, surfacing risks that human analysts often miss. While many investors experiment with AI tools for memos, sourcing, and scoring founders, they remain wary of losing the relational aspects of the business. At the same time, AI‑driven development tools enable founders to launch companies with far less capital, potentially reducing the need for large VC checks and prompting an existential debate within the industry.Lire la suite

Anthropic executives say the U.S. Department of Defense's designation of the AI startup as a supply‑chain risk has caused customers to pause or cancel deals, threatening hundreds of millions of dollars in revenue. The company has filed lawsuits alleging free‑speech violations and unfair discrimination, while major cloud providers have pledged to continue offering Anthropic’s tools outside of Pentagon work. Executives warn the fallout could undermine market confidence and jeopardize future fundraising.Lire la suite

More than 30 engineers and researchers from OpenAI and Google, including DeepMind chief scientist Jeff Dean, submitted an amicus brief backing Anthropic in its legal challenge to the Pentagon's supply‑chain risk designation. The brief argues that the blacklist threatens U.S. competitiveness in artificial intelligence and chills debate on AI safety. Anthropic seeks a temporary restraining order to keep working with military partners while the lawsuit proceeds. The filing highlights the industry's concern over unpredictable government actions and underscores the importance of contractual safeguards on AI use.Lire la suite