Actualités — Page5

Newly obtained financial statements show OpenAI’s revenue climbing from $3.7 billion in 2024 to $13.07 billion in 2025, yet research and development costs, sales and marketing spend, and compute expenses pushed its operating loss to $20.92 billion. The data, supplied to journalist Ed Zitron and reviewed by the Financial Times, arrives as the company files SEC paperwork ahead of an anticipated initial public offering, highlighting a business still far from profitability despite rapid top‑line growth.Lire la suite

AI startup Probably announced a $9 million seed investment led by Andreessen Horowitz. Founder Peter Elias says the funding will fuel development of a data‑science platform that pairs large language models with a deterministic validator, aiming for near‑perfect accuracy and the ability to run on desktop hardware. The approach promises to curb hallucinations and lower token costs for precision‑sensitive use cases such as accounting and medical analytics.Lire la suite

For the first time since its launch, OpenAI’s ChatGPT slipped below the 50% market‑share threshold, according to Sensor Tower’s State of AI Report for 2026. The shift follows a surge in Google’s Gemini and Anthropic’s Claude, which together lifted the competitive landscape to a more fragmented state. While ChatGPT still tops the global rankings with more than 1.1 billion monthly users, its share dropped to 46.4% by the end of May, reflecting a broader trend of users switching between AI assistants and a growing emphasis on monetization across the sector.Lire la suite

Anthropic executives flew to Washington, D.C., on Monday for high‑level talks with the White House and Commerce Department, but the dispute over export controls on the company’s latest AI model, Claude Fable 5, ended without a resolution. Officials still believe the model’s guardrails can be bypassed, potentially exposing powerful cybersecurity capabilities, while Anthropic maintains the risk is overstated. The impasse follows a rapid government response to alleged jailbreak vulnerabilities and adds to the firm’s ongoing battle with the Pentagon over military use of its technology.Lire la suite

The Justice Department stepped into a lawsuit filed by the NAACP over Elon Musk's xAI operating unpermitted natural‑gas turbines at its Colossus 2 data center in Southaven, Mississippi. In a filing, the DOJ argued that shutting down the turbines would jeopardize U.S. national, economic and energy security because the AI models they power support classified military missions, including recent strikes against Iran. The civil rights group contends the turbines violate the Clean Air Act and threaten public health in a region already burdened by pollution.Lire la suite

Anthropic, the maker of the Claude AI assistant, has been sued in a proposed class action alleging that its Max 5x and Max 20x subscription tiers promise usage caps that far exceed what customers actually receive. Plaintiff Karl Khan claims the $100‑ and $200‑per‑month plans fall short of advertised multipliers, prompting users to buy additional credits. The complaint, filed in the U.S. District Court for the Northern District of California, seeks class‑action status for all subscribers since April 2025 and cites damages exceeding $5 million.Lire la suite

Anthropic received a last‑minute directive from the Trump administration on Friday, ordering the company to block foreign‑national access to its newest AI models, Claude Mythos 5 and Fable 5. The move forced the San Francisco‑based lab to suspend the products it had just begun promoting and prompted senior executives to fly to Washington for emergency talks. The dispute, sparked by concerns over a reported jailbreak and possible Chinese involvement, could reshape the regulatory landscape for American AI firms.Lire la suite

OpenAI disclosed that it spent $34 billion in 2025, more than double its $13 billion revenue, as the company prepares a confidential filing for an initial public offering that could value it at up to $1 trillion. The bulk of the outlay went to research and development and sales, while a $122 billion funding round earlier this year gave the firm a valuation of $852 billion. Investors such as SoftBank and Nvidia continue to back the AI leader, even as the firm’s net loss swells to roughly $39 billion. The numbers provide a first look at the financial picture prospective shareholders would inherit.Lire la suite

London‑based Ineffable Intelligence, the startup founded by DeepMind veteran David Silver, announced a partnership with Google Cloud to run one of the largest AI clusters ever built. The deal, unveiled at Google Cloud’s London summit on June 16, will see the lab using Google’s A5X instances powered by Nvidia’s upcoming Vera Rubin GPUs. Silver said the integrated AI Hypercomputer architecture was essential for his “superlearner” research, which aims to teach machines knowledge through experience rather than static data. The move positions Ineffable at the heart of Europe’s frontier AI push.Lire la suite

French Prime Minister Sebastien Lecornu announced on Monday an extra €655 million for artificial‑intelligence projects, the centerpiece of which is a sovereign chatbot for the nation’s civil service. The digital assistant will be available to roughly one million public employees, helping them process routine paperwork, streamline judicial steps and aid researchers with grant applications. By keeping the technology under French control, the government aims to modernize administration while reducing reliance on foreign AI providers.Lire la suite

Anthropic announced it is shutting down its newest AI models, Fable 5 and Mythos 5, after the U.S. Department of Commerce imposed export controls following a reported jailbreak that could bypass the models' safety safeguards. The company said the government notified it on Friday that a vulnerability had been demonstrated, forcing a complete suspension of the services for all foreign nationals. Anthropic disputes the severity of the issue and argues the move could stall progress on frontier AI, while officials in Washington remain tight‑lipped.Lire la suite

A Washington, D.C. resident has filed a federal lawsuit accusing Anthropic of misleading customers about the usage caps of its Claude Max subscription tiers. The plaintiff, Karl Kahn, claims the Max 5x and Max 20x plans deliver far less computing power than the company advertises, prompting a request for class‑action status on behalf of other U.S. users. Anthropic has not commented on the allegations.Lire la suite

U.S. District Judge Rita F. Lin ruled on Monday that xAI's claim that OpenAI stole trade secrets is dismissed with prejudice, barring the startup from refiling. The decision follows a February dismissal and an amended complaint that still failed to show a link between OpenAI and alleged employee misappropriation. The case, first filed in September 2025, adds to a string of legal clashes between Elon Musk’s AI ventures and OpenAI, including a recent jury verdict on a separate contract dispute.Lire la suite

Anthropic took its latest AI models, Fable 5 and Mythos 5, offline over the weekend after the White House demanded the company block access for all foreign nationals, including its own staff. The abrupt move, described by Anthropic as having "little choice," triggered swift reactions in Europe and Canada, where officials warned that reliance on U.S. frontier AI poses a strategic risk. Leaders in the United Kingdom, France and Canada cited the shutdown as a catalyst for building domestic AI capabilities, while U.S. officials hinted the decision was tied to concerns about Chinese access to the models.Lire la suite

The U.S. government instructed Anthropic to block foreign access to its latest AI models, Fable 5 and Mythos 5, just days after their launch. Citing concerns that the models could be manipulated for cyber‑attack purposes, the directive forced the company to shut the services for all users, including its own staff. Anthropic complied while disputing the need for a recall. The move sparked alarm in Europe and Canada, where officials warned that reliance on U.S. frontier AI could threaten national security and urged domestic development.Lire la suite

Anthropic’s senior technical staff will sit down with U.S. Commerce Department officials on Monday in an effort to resolve the escalating dispute surrounding the government‑ordered suspension of its Fable 5 and Mythos 5 AI models. The clash began after a jailbreak vulnerability was uncovered, prompting export‑control letters and a global shutdown that has drawn criticism from cybersecurity experts, spurred a Chinese competitor’s launch, and heightened tensions between Silicon Valley and the Trump administration.Lire la suite

The U.S. Commerce Department sent Anthropic a letter on Friday invoking an export‑control directive that barred non‑American users, including the company’s own staff, from accessing its Fable 5 and Mythos 5 models. Citing an unspecified national‑security concern, the agency forced Anthropic to shut down both models for all customers. The company, unsure of the exact trigger, believes the action relates to a reported guard‑rail bypass. Security experts have condemned the move as heavy‑handed, urging the administration to rescind the order.Lire la suite

A coalition of 76 cybersecurity experts, including former Facebook security chief Alex Stamos and Bugcrowd founder Casey Ellis, has sent an open letter to the U.S. government demanding the reversal of an export‑control order that bars Anthropic’s Fable and Mythos AI models. The experts argue the ban strips defenders of powerful tools needed to discover and patch software vulnerabilities, while adversaries continue to advance. Anthropic halted worldwide access to the models after the order, citing national‑security concerns, but the signatories say the restriction is unwarranted and threatens cyber‑defense capabilities.Lire la suite

The Department of Defense’s generative AI portal, GenAI.mil, now sees 1.5 million daily users – roughly half the service’s 3.5 million workforce – after a six‑month surge sparked by the deployment of Google’s Gemini model. Undersecretary Emil Michael told a Hudson Institute event that the platform, launched in December 2025 with just 80,000 users, has become the primary AI tool for administrative tasks, congressional reporting and the creation of semi‑autonomous agents, while the Pentagon prepares a $54.6 billion budget for its Defence Autonomous Warfare Group.Lire la suite

The White House is reportedly wary that a China-linked group may have accessed Anthropic's advanced AI models Mythos and Fable, prompting the administration to tighten export restrictions on the technology. Officials have not publicly confirmed the claim, and Anthropic says the government did not raise China during its discussions. If true, the breach would represent a significant national‑security concern, as the models could be reverse‑engineered through techniques like distillation. The episode follows a prior incident where a Discord community briefly accessed Mythos before the company shut it down.Lire la suite