News

Anthropic released Claude Fable 5, its first Mythos‑class large language model, on Tuesday. The new system outperforms the company’s previous Opus line but ships with strict filters that block requests about cybersecurity, biology and chemistry. When a user asks about a prohibited topic, Fable 5 redirects the query to the older Claude Opus 4.8 model and displays a warning. Anthropic says false‑positive refusals affect fewer than five percent of sessions, a trade‑off the firm deems necessary to keep the technology from aiding malicious actors. Read more

Google unveiled Gemini 3.5 Live Translate, a speech-to-speech model that delivers real‑time translation across more than 70 languages. The new AI system, part of the Gemini 3.5 family introduced at the company’s I/O conference, promises conversational latency of only a few seconds while preserving intonation and pitch. Developers can access the model through a public preview of the Gemini Live API and AI Studio, and the feature rolls out across Google’s ecosystem, including Meet and the Translate app. Read more

Microsoft’s head of artificial intelligence, Mustafa Suleyman, walked back a February remark that AI would fully automate white‑collar work within 12 to 18 months. In a recent appearance on the Decoder podcast, Suleyman said the comment referred to specific tasks—not entire jobs—and that AI is meant to help professionals complete routine work faster. He emphasized that roles such as lawyers, accountants, and project managers will remain, even as the tools they use become more efficient. Read more

Anthropic announced Claude Fable 5, touting it as the company’s most capable model ever released to the public. The new system excels in software engineering, knowledge work and vision tasks, pulling ahead of competing models as problems grow in length and complexity. Anthropic says the launch is possible thanks to fresh safety controls that block high‑risk topics like cybersecurity and biology, with the model falling back to Claude Opus 4.8 when safeguards trigger. Pricing is set at $10 per million input tokens and $50 per million output tokens. Read more

Swedish platform Lovable announced a $500 million annualized revenue run rate, up from $400 million earlier this year, while its user base swells with non‑technical creators. The company’s first “build economy” report shows 80 percent of builders identify as non‑technical, with projects ranging from e‑commerce sites to internal CRMs generating an average of 720 million monthly visits. Lovable now serves over half of Fortune 500 firms and claims more than 50 million projects on its platform. The data, drawn from internal usage and a May‑2026 survey, is self‑reported and unaudited. Read more

Rising inference costs are prompting AI firms to reconsider the long‑standing belief that bigger models always win. Industry insiders, including Coinbase co‑founder Brian Armstrong, predict that within the next year‑and‑a‑half most workloads will migrate to less expensive models. Early tests by legal‑tech startup Harvey show a three‑fold reduction in compute spend without sacrificing quality, sparking a broader debate about the economics of large‑scale AI and the future of heavyweight labs such as OpenAI and Anthropic. Read more

Anthropic unveiled Claude Fable 5 on Tuesday, making the most powerful version of its Mythos family available to the general public for the first time. The new model, optimized for software engineering, knowledge work, and vision tasks, ships with hard safety constraints that block high‑risk queries in fields such as cybersecurity and biology. Initially offered as a preview for select partners, Fable 5 now rolls out through Anthropic’s API and enterprise plans, with a limited free‑access window that ends on June 22. The launch comes amid growing calls for industry‑wide safeguards on frontier AI systems. Read more

Anthropic announced Tuesday that its new Claude Fable 5 model is now available to enterprise customers and paid subscribers. Built on the same architecture as the restricted Mythos system, Fable 5 is the first Mythos-class model open to the public, though it comes with safety guardrails that block high‑risk queries and fall back to Claude Opus 4.8. The launch coincides with the company’s confidential IPO filing and a steep price increase, positioning Anthropic to monetize frontier AI capabilities while addressing misuse concerns. Read more

London unveiled a $1.47 billion plan to build a national AI supercomputer and fund homegrown chip makers. More than $1 billion will go toward the system, with $530 million earmarked for hardware and $200 million for specialist inference chips. The procurement strategy favors British startups such as Olix and Fractile, and researchers could access the machine by 2030. The move is part of a broader push for AI sovereignty, complementing new AI growth zones and a $675 million venture fund aimed at reducing reliance on U.S. and Asian technology. Read more

OpenAI announced it has confidentially submitted an S‑1 registration statement to the U.S. Securities and Exchange Commission, indicating the company is preparing for a potential initial public offering. While no date or price has been set, the firm said it expects the filing to leak and wants to be transparent about its plans. Valued at $852 billion after a fresh fundraising round that brought in $122 billion from investors including NVIDIA and Amazon, OpenAI reported $25 billion in annualized revenue but projects $115 billion in expenses through 2029. The move comes amid fierce competition in the AI sector and a lawsuit alleging negligence over safety‑system warnings. Read more

Google announced a major upgrade to its NotebookLM AI research tool, moving it to the new Gemini 3.5 model and adding Antigravity‑powered code execution. The refresh expands supported file types, streamlines web‑source integration, and introduces a built‑in cloud computer with more than 100 software skills. In side‑by‑side tests, the revamped NotebookLM beat its Gemini 3.1 predecessor on five core evaluation dimensions, posting a 65 percent win rate. The changes aim to cut token costs, boost accuracy and multilingual support, and let users build complex workflows without leaving the notebook environment. Read more

OpenAI announced on Monday that it has confidentially submitted a Form S‑1 to the U.S. Securities and Exchange Commission, signaling the next step toward a public offering. The filing follows Anthropic’s similar move on June 1 and comes amid internal debates over revenue targets, user growth and a massive compute budget. With valuations of $852 billion for OpenAI and $965 billion for Anthropic, the two AI firms are poised to compete for investor attention just as SpaceX prepares for what could become the largest IPO in history. Read more

OpenAI announced Monday that it has filed a confidential draft registration statement with the U.S. Securities and Exchange Commission, signalling its intent to pursue an initial public offering. The move comes just over a week after rival Anthropic filed its own confidential IPO paperwork, intensifying competition between the two AI powerhouses. While the company did not disclose pricing or timing, it said the filing gives it flexibility to go public sooner if market conditions warrant. OpenAI also released a broad statement on its mission and the future of artificial general intelligence, underscoring the high stakes as the sector gears up for a wave of high‑profile listings in 2026. Read more

OpenAI disclosed a confidential filing for an initial public offering on Monday, positioning the generative‑AI leader for one of the decade's most watched market debuts. At the same time, Sam Altman's biometric‑verification startup, Tools for Humanity, known for its Worldcoin project and silver‑orb iris scanner, is reportedly cutting jobs as revenue struggles and regulatory backlash mount. The company, once valued at $2.5 billion and backed by Andreessen Horowitz and Bain Capital, faces bans in Kenya and a fine in South Korea over privacy concerns tied to its $50‑in‑Worldcoin incentive scheme. Read more

Florida Attorney General Ashley Moody filed an 83‑page complaint on June 1 alleging that OpenAI’s ChatGPT is a dangerous product that contributed to mass shootings and suicides. The suit seeks to hold CEO Sam Altman personally liable and could bring billions in penalties. The filing follows a growing trend of product‑liability cases against AI firms, echoing legal strategies once used against social‑media giants and even the tobacco industry. More than 20 lawsuits now target OpenAI, while courts begin to reject Section 230 defenses for chatbot developers. Read more

OpenAI is overhauling ChatGPT's interface, adding new prompts that steer users toward coding tools, image generators and partner apps like Canva and Booking.com. The move signals a pivot toward business customers, as the company consolidates product teams under a single leader, phases out consumer‑centric features such as an in‑app checkout, and shuts down its short‑lived video‑generation tool, Sora. Executives say the goal is a unified AI assistant that understands intent without the need for explicit prompts, a strategy that mirrors Anthropic's recent push for profitability and an eventual IPO. Read more

Google announced a sweeping update to its AI‑driven note‑taking app NotebookLM, moving the service onto the newly released Gemini 3.5 model. The upgrade promises more accurate answers, automatic source discovery via Google Search, and the ability to run code on a secure cloud computer. Users will also see new export options, from PDFs to Excel sheets and AI‑generated images. The changes roll out first to AI Ultra and Workspace subscribers, with broader availability slated for later. Read more

Mustafa Suleyman, the head of Microsoft AI, told The Verge that the company is accelerating its superintelligence effort while expanding its model portfolio. He said the October contract with OpenAI frees Microsoft to train its own frontier models, unveiled seven new models at the Build conference, and highlighted a new partnership with the Mayo Clinic to build a health‑focused foundation model. Suleyman stressed that Microsoft’s strength lies in its enterprise reach and that the next few years will see tighter integration of custom silicon, data, and reinforcement‑learning pipelines to deliver higher‑performing AI services. Read more

By the end of the first morning of London Tech Week, the United Kingdom had locked in several billion pounds in AI investment pledges, largely aimed at expanding compute infrastructure. Prime Minister Keir Starmer unveiled a £400 million national AI compute strategy, while AMD and cloud firm Nebius together pledged roughly £3.7 billion for high‑performance computing projects. Mayor Sadiq Khan added a £12 million package to help small businesses adopt AI. The influx of capital arrives as British AI startups have already raised more than £8.2 billion this year, positioning the UK as a growing hub for European artificial‑intelligence development. Read more

OpenAI plans to roll out a major redesign of ChatGPT within weeks, adding coding tools, image‑generation capabilities and a suite of partner applications such as Canva and Booking.com. The upgrade, described as a “super app,” targets enterprise customers and aims to boost revenue ahead of a potential initial public offering. The changes will first appear on the chatbot’s website and mobile interfaces, signaling OpenAI’s push to broaden its services beyond simple conversation. Read more