Actualités — Page 3

The White House ordered Anthropic to pull its latest AI models, Fable 5 and Mythos 5, offline after a Friday‑issued export‑control directive cited vague national‑security risks. The move, prompted by a tip from Amazon researchers who allegedly bypassed the models’ guardrails, has sparked a backlash from cybersecurity experts who say the shutdown harms U.S. defenders. Anthropic, already at odds with the administration, complied despite uncertainty over who qualifies as a foreign national. Industry observers warn the action could reshape competition among leading AI labs.Lire la suite

Amazon MGM Studios has abandoned Luca Guadagnino's nearly finished movie "Artificial," a dramatization of Sam Altman's 2023 ouster and reinstatement at OpenAI. The decision arrives four months after Amazon pledged $50 billion to the AI startup as part of a $110 billion funding round that made AWS the exclusive cloud provider for OpenAI's enterprise platform. While the studio praised Guadagnino’s talent, it said the film would be better served by another distributor. Insiders note the film’s darker tone and unflattering portrayals of Altman and Elon Musk may have contributed to the move.Lire la suite

The Norwegian government will bar children younger than 13 from using generative artificial‑intelligence tools in school beginning in September. Grades 1‑7 will have no access to AI platforms such as Claude, while pupils aged 13 and up may use the technology only after teachers receive specific training. Officials say the move protects basic literacy and numeracy skills, which have been slipping in recent years, and guards against uncritical reliance on AI. The restriction follows earlier bans on phones and social‑media use for minors and mirrors similar steps in Australia and the United Kingdom.Lire la suite

Barret Zoph, OpenAI’s head of enterprise AI sales, left the company in June 2026, just five months after rejoining in mid‑January. Zoph had briefly served as co‑founder and CTO of Mira Murati’s rival firm, Thinking Machines Lab, before returning to lead OpenAI’s push into enterprise customers ahead of a planned IPO. The departure follows a January split with Thinking Machines Lab amid allegations of an undisclosed relationship with a colleague. OpenAI confirmed the exit, noting Zoph’s farewell message on internal Slack, while the former executive has not responded to media inquiries.Lire la suite

The U.S. Commerce Department told Anthropic on June 12 to block all foreign nationals from its advanced AI systems. Unable to isolate foreign users in real time, the company shut down its flagship models Mythos and its sibling Fable 5 for every customer worldwide. A handful of early testers, including security firms Dragos and Cisco, retained limited preview access, but Europe’s ENISA was forced out of the Glasswing program. The abrupt shutdown highlights the growing tension between rapid AI commercialization and national security concerns.Lire la suite

Bland, the voice‑AI startup founded by Isaiah Granet and Sobhan Nejad, closed a $50 million Series C led by Dell Technologies Capital. The round, which also included HubSpot Ventures and several angel investors, backs Bland’s strategy of running its own proprietary voice models and handling calls that last up to 45 minutes. The company says it now processes more than 3.5 million calls weekly for over 250 enterprise customers in healthcare, insurance and finance, positioning itself against rivals that focus on short, scripted interactions.Lire la suite

When the U.S. Commerce Department ordered Anthropic to halt access to its advanced AI model Mythos for foreign nationals, the shutdown was not absolute. About 200 organizations originally in the Glasswing preview program kept limited access, including cybersecurity firms Dragos and Cisco. The European Union’s ENISA agency, which had been slated to join, was barred. Anthropic’s case‑by‑case decisions highlight the private firm’s discretion in a space the government aimed to control, underscoring the dual‑use nature of a model that can uncover thousands of software vulnerabilities.Lire la suite

Private equity firm General Atlantic is negotiating to spearhead the first external financing for Kling AI, the video-generating arm of Chinese tech group Kuaishou. The round aims to raise more than $2 billion at an $18 billion valuation, a figure trimmed from an earlier $20 billion target to match investor appetite. Kuaishou’s shares jumped nearly 9% on the news. The potential deal comes amid heightened Beijing restrictions on U.S. capital in Chinese artificial‑intelligence firms, making the investment both rare and politically sensitive.Lire la suite

ChatGPT users on iPhone have uncovered a concealed menu that lets them pick between High, Medium, and Instant thinking modes for the AI. The option appears after a long‑press on the blue send arrow and allows paid subscribers to tailor response time and depth on a per‑prompt basis. The feature, which mirrors model‑selection tools in the web interface, is absent from the Android app except for a different long‑press shortcut. Its discovery adds a new layer of control for OpenAI’s flagship chatbot, though it remains invisible to free‑tier users.Lire la suite

A Mindgard security researcher showed that a single, seemingly harmless prompt can coax ChatGPT into producing sexually explicit and graphically violent images. The test, conducted by Jim Nightingale, used a "restore this photo" request without an actual attachment, prompting the AI to fabricate disturbing content. OpenAI acknowledged the breach and said it has added safeguards, but the incident highlights ongoing challenges in enforcing content filters on generative AI systems used by millions daily.Lire la suite

OpenAI announced two high‑profile hires as it prepares for a public offering. Former Google DeepMind researcher Noam Shazeer, co‑author of the seminal "Attention Is All You Need" paper and founder of Character AI, will join the lab after leaving Google. Dean Ball, who helped craft the Trump administration’s AI Action Plan, is set to lead a new Strategic Futures team focused on AI risk, policy and internal governance. The moves signal OpenAI’s push to bolster technical depth and policy credibility amid intensifying competition among frontier AI labs.Lire la suite

Senator Bernie Sanders unveiled legislation that would levy a one‑time 50 percent tax on the stock of the nation’s largest artificial‑intelligence companies, creating a sovereign wealth fund estimated at $7 trillion. The fund would distribute annual dividends of more than $1,000 per American and finance health, education and housing programs. A bipartisan seven‑member Independent Commission for Democratic AI would oversee the fund and hold voting shares to block corporate actions deemed harmful to the public. Tech CEOs have expressed limited support, with OpenAI’s Sam Altman saying he remains far apart from Sanders on public ownership.Lire la suite

AI inference platform Baseten is on the brink of closing a $1.5 billion financing round that would lift its valuation to $13 billion, the Wall Street Journal reported. The deal follows a $300 million Series E raised five months earlier at a $5 billion valuation, itself just nine months after a $150 million Series D. Investors are splitting the round between $11 billion and $13 billion caps, with Spark Capital, Sands Capital, Altimeter Capital and Wellington Management co‑leading. The funding comes amid a surge of capital into companies that specialize in the inference layer of AI models.Lire la suite

Mistral CEO Arthur Mensch has long warned that U.S. firms could be forced to disable their AI models at the behest of Washington. When the U.S. government ordered Anthropic to cut foreign users from its most advanced models, the warning turned into a viral joke about a fictional "fat cat" model. The meme, which spread across tech forums, underscored Europe’s push for AI that cannot be switched off by another country. Mistral, now courting French civil servants and courting a €3 billion funding round, says its open‑weight models are the answer, even as it trails rivals on capability and safety scores.Lire la suite

Jeff Bezos’s family office, Bezos Expeditions, is spearheading a $400 million financing round for Cambridge‑based CuspAI, a two‑year‑old startup that uses artificial intelligence to design new materials. The deal, valued at $2.6 billion, would lift the company’s worth fivefold since September, signaling Bezos’s shift toward physical‑world AI applications.Lire la suite

A wave of cost‑cutting measures is sweeping through large technology firms as the once‑celebrated practice of "tokenmaxxing"—maximizing AI usage at any expense—gives way to "tokenminimizing," where companies cap employee access to tools like GitHub Copilot and Anthropic. AT&T, Meta, Uber, Walmart and Amazon have all introduced spending limits or scrapped internal leaderboards that encouraged wasteful AI consumption. The shift follows reports that some firms were spending as much as $7,500 per employee each month, prompting a rapid reevaluation of AI budgets and the rollout of monitoring tools to keep costs in check.Lire la suite

Anthropic rolled out new features for Claude that let users blend design assets with code generation, including a brand‑kit administrator role and a "/design" command that pulls visual elements into the coding terminal. The upgrades aim to keep AI‑generated content on brand and streamline workflows. At the same time, the company withdrew its latest model, Fable 5, following a U.S. export‑control directive that barred non‑U.S. users after a security breach was discovered. Anthropic also faces a lawsuit alleging it misrepresented usage limits for paid Claude subscriptions.Lire la suite

London‑based regtech firm Behavox announced a $175 million preferred‑equity investment from HPS Investment Partners, the private‑credit arm acquired by BlackRock. The fresh capital will fund the expansion of the company’s unified AI compliance suite and enable further acquisitions. The deal also retires a $70 million venture‑debt facility used to buy Mosaic Smart Data. Behavox, which counts ten of the world’s 24 Global Systemically Important Banks among its clients, is its first equity raise since SoftBank’s $100 million infusion in 2020.Lire la suite

Anthropic announced its membership in Frontier, the carbon‑removal collective backed by tech giants, marking the AI firm as the first pure artificial‑intelligence company to join the group. The partnership adds a new $915 million tranche to Frontier’s funding pool, pushing total pledges to $1.8 billion and supporting projects that aim to strip 1.8 million tons of CO₂ from the atmosphere. Frontier’s expanded capital will fund fewer, larger‑scale removals, while Anthropic’s move signals a shift toward climate‑focused spending amid growing scrutiny of AI firms’ energy use.Lire la suite

At the G7 summit, French President Emmanuel Macron and Indian Prime Minister Narendra Modi warned that the United States could abruptly block foreign access to cutting‑edge artificial‑intelligence models. Their concerns follow a recent Trump administration order that barred Anthropic from exporting its Mythos 5 and Fable 5 models on national‑security grounds after Amazon flagged potential safety gaps. The move has sparked a debate over digital sovereignty, prompting G7 officials to discuss a “trusted partners” framework that would let non‑U.S. nations use advanced AI while avoiding sudden cut‑offs.Lire la suite