Notizie — Pagina 4

The U.S. administration instructed Anthropic to revoke South Korean carrier SK Telecom’s access to its advanced AI model Claude Mythos after concerns surfaced about the telecom giant’s alleged ties to China and Amazon’s report of vulnerabilities in a related version, Fable 5. Anthropic complied, shutting off the models for all foreign users, while both the White House and SK Telecom declined to comment on the specifics.Leggi di più

Artificial intelligence is no longer a speculative tool for movie making; it has already produced full‑length features and festival‑ready shorts. Last week, the Tribeca Film Festival showcased "Dreams of Violets," a 100% AI‑generated docudrama, while a 90‑minute thriller called "Hell Grind" was assembled in just two weeks. A comedy‑horror trailer titled "DEADLINES" on YouTube demonstrated that AI can now deliver jokes and pacing, not just visual tricks. The industry’s next hurdle is clear: convincing viewers to sit down and watch AI‑crafted stories.Leggi di più

ByteDance, the owner of TikTok, is on track to spend more than $1 billion each year on Microsoft’s AI and cloud services, making it the tech giant’s biggest AI customer. The Chinese firm primarily buys OpenAI models through Azure, a channel that OpenAI and Anthropic avoid for Chinese clients. While the partnership fuels Microsoft’s fastest‑growing AI market, it also sits uncomfortably amid rising U.S. concerns over China’s AI ambitions and new regulatory limits on model access.Leggi di più

Yann LeCun, founder of AMI Labs and former Meta chief AI scientist, told CNBC that Elon Musk’s xAI is “kind of a failure,” citing staff turnover, hiring challenges and costly infrastructure. He warned that rising AI service costs and thin profit margins could trigger a bubble, echoing concerns from other industry leaders. LeCun’s critique doubles as a promotion for his own “world‑model” approach, which his startup AMI Labs has raised about $1 billion to develop.Leggi di più

Midjourney, the AI image‑generation firm, announced its first hardware product: a full‑body ultrasonic scanner that promises a 3‑D body map in under a minute. Developed with Butterfly Network’s ultrasound‑on‑chip technology, the device will be housed in consumer‑focused spas, starting in San Francisco next year. The company says the scan matches MRI detail while cutting scan time from 60‑90 minutes to about 60 seconds. Midjourney plans to seek FDA clearance within the year and aims to deploy 50,000 units worldwide by 2031, hoping to slash deaths and health‑care costs.Leggi di più

Anthropic’s Claude lets users import and export conversational memory without starting from scratch. The process involves a few clicks in the Settings menu, where you can pull data from other AI providers or download a ZIP file of your Claude history. While Claude streamlines memory transfers, it still lacks a direct upload feature for external ZIP files, and other chatbots offer comparable but sometimes more cumbersome options.Leggi di più

Microsoft is evaluating a self‑hosted, fine‑tuned version of the open‑source Chinese AI model DeepSeek V4 – or another open‑source alternative – as a lower‑cost engine for its enterprise Copilot Cowork assistant. The move comes as the company shifts the product to usage‑based pricing and seeks to temper soaring compute costs tied to repeated model calls. Microsoft says any new model would run on Azure, remain optional for customers, and include safeguards to address bias and security concerns.Leggi di più

Anthropic has rolled out a major update to its Claude Design assistant, adding the ability to work directly from a local codebase and to hand off projects to Claude Code without starting from scratch. The refresh also brings a flexible import tool for design systems, a refined image editor, new admin controls, and shared usage limits across Anthropic’s AI products. With more than a million users in the first week of its preview, the company signals that Claude Design is moving out of beta even as it faces a lawsuit over its Max‑plan limits.Leggi di più

OpenAI rolled out a new Scheduled page for ChatGPT on June 17, 2026, giving paying subscribers a dedicated hub to queue, edit, pause or delete future prompts. Users can pick exact times or broader windows—morning, afternoon, evening—to run tasks, and the company says the feature is faster and more reliable. The update arrives for Go, Plus, Pro, Business and Enterprise plans on both web and mobile, while the free tier remains excluded. At the same time, OpenAI is retiring Pulse, its daily summary tool, with a short transition period for pro users.Leggi di più

Google has set a June 25 release date for its first smart speaker in nearly six years, pricing the new device at $100 and opening preorders today. The round, fabric‑covered speaker—available in four colors—offers 360-degree sound, a glowing light ring that indicates listening status, and a quad‑core processor with a dedicated NPU for on‑device AI. Three far‑field microphones and a mute switch give users control over voice activation, while capacitive touch controls manage playback without a spoken command.Leggi di più

Anthropic unveiled Mythos Preview in April, showcasing an AI model whose creative and tenacious capabilities could be weaponized for cyber‑attacks within months. Industry experts, including TPO Group’s chief security officer Tarah Wheeler and Veracode co‑founder Chris Wysopal, warn that rival firms and open‑source projects are likely to produce comparable tools, amplifying the risk. OpenAI’s own cybersecurity‑focused model, released a few weeks earlier, underscores a broader shift toward AI‑driven hacking. Leaders in the security community have urged the White House to rethink its export‑control measures, arguing that blanket restrictions may hamper defensive innovation without curbing threats.Leggi di più

Anthropic spent the week scrambling to restore access to its flagship Fable 5 and Mythos 5 chatbots after the Trump administration ordered a blanket cut of service for all foreign nationals, including U.S. users and the company’s own staff. The government cited national‑security concerns and an export‑control directive, but offered no public legal justification. Experts say the move is unprecedented, exposing gaps in how U.S. export rules apply to cloud‑based AI and raising questions about the future of frontier‑model governance.Leggi di più

At a closed‑door lunch on the final day of the G7 summit in Évian‑les‑Bains, Anthropic CEO Dario Amodei and DeepMind chief Demis Hassabis urged the United States to spearhead an international AI coalition. Canadian Prime Minister Justin Trudeau’s team signaled support, while OpenAI’s Sam Altman called for a global forum to set testing standards. Attended by roughly a dozen tech leaders and G7 heads of state, the meeting produced no binding agreements but highlighted growing friction between Washington’s export controls and the AI industry’s push for coordinated governance.Leggi di più

A Pew Research Center survey finds that just 16 percent of U.S. adults expect artificial intelligence to have a positive impact over the next two decades, while 40 percent foresee a negative effect. The study also reveals widespread skepticism about government regulation and corporate safety, especially among younger adults. At the same time, AI chatbots are becoming a daily tool for a quarter of the population, with ChatGPT leading usage at 44 percent. Gender and age gaps shape both attitudes and adoption rates.Leggi di più

Anthropic announced a pause on its planned token‑based billing changes for the Claude Agent SDK, citing feedback from developers who warned of steep cost increases. The company updated its billing support page on Monday, stating that nothing has changed for now while it works on a new pricing structure. The move follows similar pricing shifts at GitHub Copilot and comes as Anthropic prepares a potential IPO, having filed confidential paperwork with the SEC. Anthropic executives emphasized the need to align subscription plans with the heavy usage patterns of third‑party tools.Leggi di più

Anthropic topped OpenAI in May for the first time, capturing 41% of business AI subscriptions according to Ramp data. The AI lab closed a $65 billion funding round, filed confidential IPO paperwork and launched Opus 4.8, even as the Trump administration ordered a ban on non‑American access to its latest models, Mythos 5 and Fable 5. Despite the government’s move, corporate demand for Anthropic’s Claude models kept rising, suggesting the controversy may be boosting the company’s market profile.Leggi di più

Anthropic took its latest AI models, Claude Fable 5 and Mythos 5, offline this week after the White House issued an export‑control order barring foreign nationals from accessing the services. The directive, announced Friday, cites the models’ ability to locate software vulnerabilities and devise exploits as a national‑security risk. Anthropic, which has been negotiating with officials since the order’s release, says the models were launched under strict guardrails but could be stripped of those protections. Industry experts warn the move signals a broader clash between rapid AI advances and government regulation.Leggi di più

OpenAI’s ChatGPT subscriptions have stayed at $20 for Plus and $200 for Pro since their launch, even as the chatbot’s capabilities have expanded dramatically. A growing chorus of users now argues the service is underpriced, pointing to the massive compute power and data‑center expenses required to run the models. OpenAI executives have acknowledged that the Pro tier loses money, fueling the debate about whether the current pricing reflects true costs or a market‑share strategy.Leggi di più

SpaceX announced an all‑stock deal to buy AI startup Cursor for $60 billion, finalizing a partnership first outlined in April. Cursor, known for its eponymous AI‑driven coding assistant, had been courting $2 billion in fresh capital from investors such as Andreessen Horowitz and NVIDIA, though that money would not have covered its operating shortfall. The acquisition gives SpaceX a foothold in the artificial‑intelligence arena as it seeks to revamp its troubled xAI division, which has faced ethical controversies and a mass exodus of senior staff.Leggi di più

SpaceX announced an all‑stock transaction to buy AI‑driven coding tool Cursor for $60 billion, with the deal slated to close in the third quarter. The acquisition follows SpaceX’s recent IPO and its merger with xAI, marking a major restructuring move for the aerospace firm. Cursor, a Visual Studio Code‑based IDE that embeds large‑language‑model features, has seen rapid revenue growth but faces stiff competition from rivals such as Anthropic’s Claude Code. The deal gives Cursor access to xAI’s compute resources and joint model‑training initiatives.Leggi di più