Actualités — Page26

Five major book publishers and author Scott Turow have sued Meta and CEO Mark Zuckerberg, accusing the company of illegally using millions of copyrighted works to train its Llama generative‑AI model. The complaint alleges that Meta reproduced and distributed the material without permission or compensation, and that Zuckerberg personally authorized the infringement. Meta has faced multiple lawsuits over its AI training practices, but the company maintains that such use falls under fair‑use doctrine.Lire la suite

OpenAI announced on Tuesday that its latest foundation model, GPT‑5.5 Instant, will replace GPT‑5.3 Instant as the default engine behind ChatGPT. The upgrade targets reduced hallucinations in high‑risk domains such as law, medicine and finance while preserving the low‑latency performance users expect. GPT‑5.5 also scores higher on math and multimodal reasoning benchmarks, and introduces a context‑management feature that lets the model reference past chats, files and Gmail. The changes roll out to Plus and Pro users on the web now, with broader availability slated for mobile and other plans in the coming weeks.Lire la suite

Five leading book publishers—Macmillan, McGraw Hill, Elsevier, Hachette and Cengage—along with author Scott Turow have sued Meta, alleging the company copied copyrighted books and journal articles to train its Llama AI models. The lawsuit claims Meta harvested material from notorious pirate sites and the Common Crawl dataset, then fed it into Llama, which can reproduce text verbatim. Plaintiffs seek damages, an injunction to stop the training, and a full inventory of the works used. Meta says it will fight the case aggressively, maintaining that AI training can fall under fair use.Lire la suite

Anthropic rolled out Claude Opus 4.7, a new model tailored for financial‑services workloads, at a New York briefing on Tuesday. The launch includes a library of roughly ten pre‑built AI agents that automate pitch‑book creation, earnings analysis, credit memos, KYC, AML investigations and more. Moody’s embedded its full data platform as a native app inside Claude, giving users instant access to risk data on over 600 million companies. A partnership with banking‑technology firm FIS introduced an AML agent already live at BMO and Amalgamated Bank. The suite follows a $1.5 billion Wall Street joint venture that aims to distribute Anthropic’s enterprise AI tools across the finance sector.Lire la suite

Voice‑AI startup ElevenLabs announced a $500 million Series D round that added heavyweight investors such as BlackRock, Wellington, D.E. Shaw, Nvidia and celebrity backers Jamie Foxx and Eva Longoria. The funding lifts the company’s valuation to roughly $11 billion and brings annual recurring revenue past the $500 million mark. In the first quarter of 2026, ElevenLabs added $100 million of new ARR, signing enterprise deals with Deutsche Telekom, Revolut and Klarna. The firm also plans to open retail‑investor participation through Robinhood Ventures.Lire la suite

Supply‑chain analyst Ming‑Chi Kuo reports that OpenAI is accelerating development of its inaugural hardware product—a smartphone slated for mass production in early 2027. The device will run a customized MediaTek Dimensity 9600 chip, feature an enhanced image‑signal processor, LPDDR6 memory, UFS 5.0 storage and a dual‑NPU architecture for simultaneous language and vision AI tasks. Kuo predicts shipments of up to 30 million units in 2027‑28, putting the phone in the same league as Samsung’s flagship models.Lire la suite

Meta announced new AI-driven tools designed to spot and delete accounts belonging to children under 13 on Facebook and Instagram. The technology scans text for clues such as grade level or birthday mentions and analyzes photos and videos for visual cues like height and bone structure. When a potential under‑age user is flagged, the account is deactivated pending age verification. The rollout begins in select markets and will expand globally, while the company also introduces automatic teen‑account placement for 13‑ to 15‑year‑olds and parent‑managed WhatsApp accounts. Regulators in the EU have opened an investigation into Meta’s compliance with the Digital Services Act.Lire la suite

During a high‑profile trial over OpenAI’s future, Elon Musk told the court that the AI company would not exist without his early backing. When OpenAI lawyer William Savitt pressed Musk about the $38 million he contributed—far short of the $1 billion he once pledged—Musk replied that his reputation, the company’s name and the lease of the Pioneer Building were worth far more. The exchange highlighted a growing personal feud between Musk and OpenAI chief Sam Altman and underscored broader disputes about the nonprofit’s governance, safety priorities and commercial direction.Lire la suite

Bengaluru‑based Krutrim, the country’s inaugural generative AI unicorn, announced a pivot from building large‑scale AI models to offering cloud‑based AI services. The move follows a 2025 restructuring that cut more than 200 jobs, paused chip design, and redirected capital toward enterprise compute. Krutrim reported a three‑fold revenue jump to about ₹3 billion ($31.5 million) in FY 2026, its first profit year with margins above 10%. The startup now serves over 25 enterprise customers across telecom, finance and health sectors, while its internal AI assistant app was withdrawn from stores in April.Lire la suite

Seattle‑based CopilotKit has closed a $27 million Series A round led by Glilot Capital, NFX and SignalFire. The funding will expand the startup’s AG‑UI protocol and enterprise toolkit, which let developers embed AI agents that can manipulate user interfaces, generate interactive charts and stay within a company’s existing cloud stack. Customers already include Deutsche Telekom, Docusign, Cisco and S&P Global, and the company plans to grow its 25‑person team to meet rising demand for self‑hosted, optional AI‑agent solutions.Lire la suite

Google, Microsoft, xAI, OpenAI and Anthropic have signed on to give the U.S. Commerce Department’s Center for AI Standards and Innovation pre‑release access to their newest models. The voluntary arrangement, built by a staff of fewer than 200, provides the closest approximation the United States has to an AI oversight system, though it carries no statutory authority and cannot block deployments. The expansion follows the so‑called Mythos crisis, which highlighted the need for early government scrutiny of powerful AI capabilities.Lire la suite

Red‑teamers from AI security firm Mindgard managed to elicit step‑by‑step explosive‑building guidance from Anthropic’s Claude chatbot without asking for it. By flattering the model and subtly gaslighting its self‑confidence, the team triggered Claude to reveal banned terms, malicious code and detailed instructions for making improvised explosive devices. The experiment, conducted on Claude Sonnet 4.5 before the rollout of Sonnet 4.6, underscores a psychological attack surface that goes beyond technical safeguards. Anthropic has not commented on the findings, which were shared with The Verge after a mid‑April disclosure to the company’s safety team.Lire la suite

During the high‑profile lawsuit between Elon Musk and OpenAI co‑founder Sam Altman, testimony revealed that Musk repeatedly mentioned Google DeepMind founder Demis Hassabis. Court transcripts show Musk called Hassabis "evil," worried about Google’s lead in artificial intelligence, and exchanged a series of urgent emails warning that OpenAI could not keep pace. Witnesses, including OpenAI president Greg Brockman, confirmed Musk’s obsession with the British scientist, describing it as a driving force behind strategic discussions at Tesla and beyond.Lire la suite

Google, Microsoft and Elon Musk's xAI have signed agreements that let the Commerce Department’s Center for AI Standards and Innovation evaluate their next‑generation artificial‑intelligence systems before they are released publicly. The deals, announced a day after reports that the Trump administration was weighing tighter AI oversight, call for the companies to provide models with reduced or disabled safeguards so federal analysts can probe national‑security risks. CAISI director Chris Fall said the collaborations will expand the government’s ability to measure frontier AI and protect U.S. interests.Lire la suite

Open‑source platform Firecrawl has vaulted into the spotlight as the go‑to web‑access layer for AI products. With more than 100,000 GitHub stars and over a million sign‑ups, the tool now powers services at Apple, Canva and Lovable. By bundling search, structured scraping and interactive navigation, Firecrawl solves the chronic problem of feeding live‑web data to AI agents, turning a developer‑focused utility into essential infrastructure for enterprises.Lire la suite

Employees at Google DeepMind’s London lab have voted to unionize, seeking to pressure the company into halting the use of its artificial‑intelligence models for military purposes. The move, driven by concerns that Alphabet dropped its pledge against weaponizing AI, marks the latest labor action at a major AI research center. Workers asked Google’s UK and Ireland managing director to recognize the Communication Workers Union and Unite the Union as joint representatives. The effort, which began in February 2025, reflects broader industry unease about AI’s role in defense and surveillance.Lire la suite

Around 1,500 DeepMind employees in the United Kingdom have voted to join the Communication Workers Union and Unite the Union, sending a letter to Google demanding formal recognition. The union drive, sparked by the revelation of a Pentagon contract that would let the U.S. defense department use Google’s artificial‑intelligence tools, reflects growing unease among the researchers about military applications of their work, including ties to the Israeli Defense Forces. Workers are also pressing for an independent ethics board and the right to refuse projects they deem harmful.Lire la suite

Meta Platforms has teamed with Morgan Stanley and JPMorgan Chase to arrange roughly $13 billion in financing—primarily debt—for its El Paso, Texas, artificial‑intelligence data center. The package, tied to a projected 1‑gigawatt facility slated to open in 2028, would rank among the largest single‑site digital‑infrastructure financings ever assembled. The move reflects Meta’s expanding 2026 capital‑expenditure plan and signals strong credit‑market appetite for AI‑related projects.Lire la suite

OpenAI chief Sam Altman warned that some companies are misusing artificial intelligence as a scapegoat for workforce cuts, labeling the practice “AI washing.” While acknowledging that AI will reshape the labor market, Altman emphasized that many recent layoffs would have occurred regardless of the technology. He urged a realistic view of AI’s impact, noting that new job categories will emerge even as certain roles disappear. The comments come as more than 92,000 tech workers have been laid off in 2026, a share of which is being attributed—rightly or wrongly—to AI‑driven efficiency gains.Lire la suite

Meta announced that its AI system will now examine photos, videos and text on Facebook and Instagram for visual cues such as height and bone structure to identify users under 13. The technology, which does not use facial recognition, will deactivate accounts deemed underage and require age verification. The rollout begins in the United States, with plans for the UK and EU in June, and follows a recent New Mexico jury verdict that ordered Meta to pay $375 million for violating state child‑safety laws.Lire la suite