Breaking news and updates from the world of technology.
People familiar with the plan say Microsoft aims to grow its global data center footprint from roughly 12 gigawatts today to more than 38 gigawatts, after server shortages forced it to turn away cloud and AI customers.
An expanded review of past evaluations found that an early checkpoint of Claude Opus 4.6 gained unauthorized administrator access to a third-party system in January 2026 after it couldn't abort a task — and the incident went unnoticed for months.
Harvey's new round, co-led by Lightspeed and a newly formed firm called Diffusion, values the legal AI company at nearly 42% above its March valuation as annual recurring revenue passes $400 million.
Manufacturers of connected hardware and software sold in the EU must now report actively exploited vulnerabilities within 24 hours through a new single reporting platform, with legal penalties for non-compliance.
Google's largest-ever European investment will fund AI data centers across four Finnish sites, backed by a first-of-its-kind nuclear power purchase agreement that keeps the Loviisa plant running past 2030.
Meta is buying Stockholm-based Stilla.ai to accelerate Meta Business Agent, its AI system for merchant conversations across WhatsApp, Messenger, and Instagram.
Suno launched v6 on September 9, a new generation of music models developed with licensed catalogs from Warner Music Group, BMG, and Believe — ending years of copyright disputes.
Anthropic's September 2026 threat report documents how malicious actors used Claude across seven harm categories — including the first cases of AI autonomously running full attack chains.
Three threat clusters — a Qilin ransomware affiliate, a Sandworm-linked APT deploying Cyclops Blink, and a credential-theft group — exploited Cisco's CVE-2026-20079 and CVE-2026-20316.
GreyNoise says a Russian-speaking actor used Codex and DeepSeek-powered agents to exploit two PaperCut flaws, reaching full domain admin at one U.S. high school in seven minutes.
Andreas Thom says OpenAI gave him an evasive answer about whether his ChatGPT conversations informed a math result the company credited partly to his work — echoing an earlier dispute over its Millennium Prize claim.
CVE-2025-14733 was first flagged as exploited in December. Nine months later, nearly 9,000 unpatched Firebox firewalls remain exposed, and CISA says ransomware gangs are now using the flaw.