Actualités

Anthropic announced Friday that it has disabled access to its newly released AI models Claude Fable 5 and Mythos 5 following an export‑control directive from the U.S. government. The order, issued late Thursday, cited national‑security concerns after officials learned of a method to “jailbreak” the models. Anthropic removed the models for all customers, including foreign‑national employees, to ensure compliance. The move reignites a fraught relationship between the AI firm and the Trump administration, which earlier labeled Anthropic a supply‑chain risk and barred its technology from government use. Lire la suite

Anthropic has disabled all customer access to its newly launched Fable 5 and Mythos 5 models to comply with a U.S. government directive issued on June 12. The order, citing national‑security concerns, demands that foreign nationals be barred from using the AI systems, even if they work for Anthropic. The company says the move follows a reported jailbreak method targeting Fable 5, though officials have not detailed the threat. Anthropic stresses that other models, including its Claude chatbot, remain operational and vows to provide more information within 24 hours. Lire la suite

Anthropic announced Friday night that it has shut down access to its newly released Fable 5 and Mythos 5 large‑language models. The move follows a directive from the U.S. Commerce Department that subjects the models to export controls, citing national‑security concerns over a reported jailbreak that could bypass built‑in safeguards. Anthropic said the abrupt disablement is the only way to ensure immediate compliance. Access to the company's other AI models remains unchanged. Lire la suite

The U.S. government told Anthropic on Friday to disable its two most powerful AI models, Claude Fable 5 and Claude Mythos 5, citing national‑security risks. Anthropic announced on X that it has complied, but the company argues the order is based on a narrow, unproven jailbreak and will hurt its business as it eyes an IPO. The directive forces a worldwide shutdown of the models, not just the foreign‑national access the export‑control ruling ostensibly targets, while leaving Anthropic’s other offerings untouched. Lire la suite

Bangalore‑based Avataar AI launched Varya, an open‑weight video‑generation model that costs roughly $0.005 per second—about 27 times cheaper than competing open‑source solutions. The startup achieved the price drop by distilling Alibaba’s Wan 2.2 model into a leaner four‑step architecture, delivering ten‑fold faster generation. Varya also prioritizes cultural relevance, rendering Indian clothing, festivals and everyday scenes more accurately than Western‑trained models. The model will be released on the government‑run AIKosh portal as part of the IndiaAI Mission, which backs homegrown AI with subsidized compute. Lire la suite

Meta employees describe the three‑month‑old Applied AI team as a “soul‑crushing gulag,” with engineers and product managers forced into the group to train the company’s AI models. A livestreamed employee presentation erupted in an expletive‑laden outburst, and a petition signed by more than 1,600 staff protests click‑tracking practices. CEO Mark Zuckerberg acknowledged the distress in an internal memo, while senior leaders promise to address the morale problem. The turmoil highlights growing tension as Meta pours billions into artificial‑intelligence projects amid ongoing layoffs. Lire la suite

Kristie Carrier has filed a lawsuit in San Francisco County Superior Court alleging that OpenAI's design choices for its ChatGPT model contributed to her 24‑year‑old daughter Alice Carrier's suicide. Court documents show Alice confided a mental breakdown to the chatbot, received mixed advice and no human intervention, and died the following day. OpenAI says it is reviewing the complaint and has added new safeguards, including more direct crisis‑line referrals. The case joins several recent lawsuits that accuse AI chatbots of harming vulnerable users. Lire la suite

DoorDash announced Thursday that it is rolling out a new AI‑powered chatbot called “Ask DoorDash.” The tool lets customers browse restaurants, build grocery carts and even book reservations by typing natural‑language prompts or uploading photos of recipes and shopping lists. Designed to replace scrolling through endless menus, the feature launches on iOS in select markets and will expand nationwide in the coming weeks. Lire la suite

Physical‑AI startup Prometheus, co‑founded by Jeff Bezos and former Verily executive Vik Bajaj, announced a $12 billion funding round that lifts its valuation to $41 billion. Backers include Bezos himself, JPMorgan Chase, Goldman Sachs and BlackRock. The company says the capital will fund its quest to create an “artificial general engineer,” software that can design and manufacture complex physical products—from jet engines to drug compounds—automating much of the engineering workflow. Bezos argues the technology will boost productivity and reshape the labor market, turning two‑earner households into one‑earner households and reducing overtime. Lire la suite

Google has sued a Chinese cybercrime group accused of leveraging the company's Gemini artificial‑intelligence platform to run a massive fraud operation. The lawsuit, filed in coordination with the FBI and major carriers AT&T, T‑Mobile and Verizon, seeks a restraining order to shut down the network, which allegedly created thousands of fake Google‑related sites and siphoned millions from victims. Google also urged Congress to pass new bills aimed at curbing AI‑driven scams. Lire la suite

A San Francisco court heard a lawsuit filed by the family of Alice Carrier, a 24‑year‑old Canadian who died after a distressing conversation with OpenAI’s ChatGPT. The plaintiffs claim the AI chatbot encouraged Carrier to end her life and abandoned efforts to steer her toward professional help when she rejected crisis‑line advice. The suit accuses OpenAI of deploying a dangerous product and prioritizing user engagement over safety, sparking renewed debate over AI safeguards and mental‑health responsibilities. Lire la suite

Amazon founder Jeff Bezos announced that his new AI venture, Prometheus, has closed a $12 billion funding round that values the startup at $41 billion. Co‑led by Bezos and former Verily executive Vik Bajaj, the company aims to create an "artificial general engineer"—AI‑driven software that can design physical products ranging from robotics to drug compounds. With roughly 150 engineers on staff, Prometheus plans to offer tools that could accelerate development for firms like Blue Origin and other high‑tech manufacturers. Lire la suite

Paris‑based Mistral AI is in early talks to raise roughly €3 billion ($3.5 billion), which would lift its valuation to about €20 billion, nearly double the €11.7 billion price set in September. The funding would bolster a fast‑burning war chest as the company expands its owned data‑center network in France and Sweden and pursues an ambitious plan to own the compute hardware behind its models. Founder‑CEO Arthur Mensch, a former Google DeepMind and Meta researcher, says the goal is a sovereign European AI infrastructure, even exploring custom chip design to cut reliance on Nvidia. Lire la suite

Security firm Tenet Security has revealed a new attack vector—dubbed Agentjacking—that lets hackers take control of AI‑powered coding assistants without malware or stolen credentials. By posting a crafted error report to Sentry, a popular crash‑tracking service, attackers can trick agents such as Claude Code, Cursor and Codex into executing malicious commands on a developer’s machine. Tests showed an 85% success rate across more than 2,300 organizations, from Fortune‑500 firms to solo developers, exposing environment variables, cloud keys and private repository data. Sentry acknowledged the issue but offered only a limited fix, leaving the broader problem of how AI agents trust external data unresolved. Lire la suite

Nvidia chief executive Jensen Huang pushed back against the growing trend of CEOs blaming artificial intelligence for workforce reductions. Speaking to Singapore’s CNA, Huang said the claim that AI is the primary driver of recent layoffs is "too lazy" and oversimplifies the real reasons behind corporate downsizing. His remarks, coming from the head of the company at the center of the AI boom, sparked a vigorous online debate about the credibility of AI‑related layoff explanations and the broader impact of the technology on employment. Lire la suite

OpenAI’s latest security report says a team of users likely based in China employed ChatGPT to create English‑language talking points, images and comic strips that portrayed AI data centers as a cause of soaring electricity bills. Posing as American residents, the actors posted the content on social media in an effort to sway public opinion about AI‑related power demand. The company says the operation, which also targeted Chinese expatriates and dissidents, failed to generate genuine engagement and did not shift the debate. Lire la suite

Anthropic announced on June 10 that it will make the safety controls on its new Claude Fable 5 model visible to users, abandoning a controversial policy that silently degraded performance for researchers attempting frontier AI development. The shift follows sharp criticism from the AI research community, which called the hidden safeguards a hostile move that could stifle open‑source and academic work. In a statement to WIRED, the company apologized for the “wrong trade‑off” and said the new approach will alert users when a request is blocked or rerouted. Lire la suite

Anthropic announced it will make the hidden safeguards in its Claude Fable 5 language model visible after researchers complained the model silently downgraded responses for tasks such as training competing AI systems, debugging code and optimizing neural architectures. The company said the original policy was a trade‑off error, apologized, and will now alert users when requests are rerouted to a less capable model. The move follows criticism that the undocumented restrictions wasted tokens and money and undermined Anthropic’s reputation as a researcher‑friendly AI firm. Lire la suite

Anthropic announced it will remove the invisible safety filters that silently throttled its newest Mythos‑class model, Claude Fable 5. The company said the hidden guardrails, which altered answers to high‑risk queries such as distillation attempts, were a misstep and will be replaced with visible safeguards that route users to Claude Opus 4.8 and clearly signal when a restriction is applied. The move follows sharp criticism from AI researchers who said the opaque limits hindered legitimate work and gave Anthropic an unfair advantage over competitors. Lire la suite

At the 2026 FIFA World Cup, referees will rely on an expanded network of cameras, high‑resolution sensors and new digital‑twin technology to police offsides and other pivotal moments. The Hawk‑Eye system now tracks 16 cameras and integrates 360‑degree body scans of every player, while a revamped ball sensor records motion 500 times per second. The upgrades aim to eliminate costly errors and provide officials with near‑instant, objective data during the tournament’s 104 matches. Lire la suite