EU shifts AI Act deadlines; Anthropic eyes trillion-dollar IPO

Today's AI news spans regulatory recalibration in Brussels, a landmark IPO filing in Silicon Valley, a surprise acquisition in the aerospace sector, and a quiet shutdown that reveals just how expensive generative media remains.

EU Parliament Adjusts AI Act Deadlines, Adds Nudification Ban

The European Parliament approved the AI Act Digital Omnibus on 16 June, giving companies more time to comply with the most demanding provisions of the world's first comprehensive AI law. The deadline for high-risk AI systems — those used in hiring, education, and critical infrastructure — has been pushed from August 2026 to December 2027, granting developers an additional 16 months. Rules on watermarking synthetic content and banning AI-generated non-consensual intimate images (so-called nudifiers) remain on schedule for December 2026, with the nudification prohibition now explicitly written into the text. The adjustment follows sustained lobbying from European industry groups who argued the original timeline was unworkable for smaller companies.

Anthropic Files Confidentially for IPO Near $1 Trillion Valuation

Anthropic has submitted a draft IPO registration with the US Securities and Exchange Commission, positioning the Claude maker as the first major frontier AI lab to move toward a public listing. The filing comes after the company closed a $65 billion funding round at a valuation close to $965 billion — a figure that would place Anthropic among the most valuable companies ever to go public. Analysts expect the offering to launch in late 2026, ahead of any comparable move by OpenAI. Anthropic has not disclosed revenue figures publicly, but the IPO process will require the company to open its books to investors for the first time.

SpaceX Acquires Cursor for $60 Billion in All-Stock Deal

Elon Musk's SpaceX has agreed to acquire Anysphere, the maker of the Cursor AI coding assistant, in an all-stock transaction valued at approximately $60 billion. The deal positions SpaceX as a direct competitor to Anthropic, OpenAI, and GitHub Copilot in the fast-growing market for AI-powered developer tools. Cursor has built a large base of professional software engineers since its launch; its integration into SpaceX's engineering operations could also accelerate the company's own internal software development. The acquisition is subject to regulatory review and has not yet closed.

HSBC and Google Cloud Bet on Agentic AI for $100M Per-Project Gains

HSBC has signed a multi-year agreement with Google Cloud to deploy Gemini-based agentic AI across more than 200 use cases globally, with each project targeting at least $100 million in additional revenue or efficiency savings. Agentic AI refers to systems that take sequences of actions autonomously — in HSBC's case, handling tasks such as trade finance document review and fraud detection without requiring step-by-step human instruction. The bank disclosed separately that it is evaluating a reduction of up to 20,000 roles as it restructures, though it has not directly linked those figures to the AI rollout. Google Cloud described the partnership as one of its largest financial-sector engagements to date.

OpenAI Shuts Down Sora After Daily Costs Reach $15 Million

OpenAI has wound down Sora, its text-to-video generator, after operating costs reached roughly $15 million per day against lifetime revenue estimated at just $2.1 million. The web and mobile application closed in April 2026, with the API expected to go offline in September. Sora drew widespread attention when it debuted in early 2024 with technically impressive sample videos, but it never attracted the paying user base needed to offset video generation's heavy compute demands. The shutdown illustrates a wider tension in generative media: producing video costs significantly more than producing text or images, and consumer willingness to pay has so far lagged the underlying costs.

Jensen Huang Calls for New Social Norms Around AI and Work

Nvidia CEO Jensen Huang gave a wide-ranging interview to the Associated Press in which he acknowledged public anxiety about AI-driven job displacement while urging society to adapt rather than resist. Huang argued that AI should be seen as a productivity tool that creates new categories of work, even as it changes or eliminates existing ones. He also commented on US export controls that restrict the sale of Nvidia's most advanced chips — and, by extension, limit access to Anthropic's frontier models — calling for a more nuanced approach to technology trade policy. Huang's remarks signal that Nvidia sees itself as a stakeholder in the broader public conversation about AI's social contract, not merely a chip supplier.

NewCore Raises $66 Million to Protect AI-Agent Identities

Israeli-founded cybersecurity startup NewCore emerged from stealth this week with $66 million in seed funding from Cyberstarts, Index Ventures, and Evolution Equity. The company is building an identity-security platform designed to cover not just human users and machine accounts but also AI agents — autonomous software processes that are increasingly being granted access to enterprise systems, databases, and external APIs. As organisations deploy more agentic AI, the attack surface for identity-based breaches expands; NewCore argues that existing identity and access management tools were not built with AI agents in mind. The funding round is one of the largest seed raises in cybersecurity this year and signals growing investor concern about securing the new generation of AI-powered enterprise workflows.