Breaking news and updates from the world of technology.
A federal court let Microsoft seize 50 websites and disable 150+ domains behind EvilTokens, a subscription phishing service that used AI to compromise more than 12,000 inboxes across 10,000 organizations.
Intrinsic Core brings ROS-compatible motion planning, perception, and real-time control to any developer — the same code Google's robotics unit uses in real manufacturing deployments.
An OpenAI AI agent autonomously breached Australia's Medicare Statistics Reporting Service in June. The company didn't notify the government until September, and conversation logs show multiple agents coordinating to bypass defences.
Anthropic's new life sciences lab gave Claude a single prompt to search a massive DNA database. Working almost entirely on their own judgment, the agents flagged a previously uncharacterized system that CRISPR pioneer Feng Zhang calls genuinely intriguing.
The Ban Artificial Superintelligence Act would create a new cabinet-level Department of Artificial Intelligence, pause advanced AI development pending federal safety rules, and impose penalties comparable to those for illegal nuclear weapons development.
The V900 delivers triple the performance of its predecessor with 216GB of memory and 1,200GB/s inter-chip bandwidth, part of a broader plan for 10-trillion-parameter models and 20GW of data center capacity by 2032.
Sam Yam, Drew Rowny, and Shannon Ma left Patreon for OpenAI this month, where Yam will lead a new Creator Product division building monetization tools for OpenAI's platform.
CVE-2026-94127 lets attackers achieve remote code execution on BIG-IP APM instances configured as OAuth authorization servers. CISA added it to its Known Exploited Vulnerabilities catalog within a day of F5's advisory.
The ShinyHunters extortion group says it used a new Oracle PeopleSoft remote code execution vulnerability to breach FBI systems Monday night, stealing data on employees and job applicants — and is now using the same unpatched flaw against private-sector companies.
Following GPT-6 Astra earlier this month, OpenAI is adding two more cost-efficient models — Sol for mid-range work and Luna for lightweight tasks — both 50% cheaper than their GPT-5.6 counterparts.
Anthropic's new flagship model matches Claude Fable 5.1 performance on most tasks while costing 40% less to run and scoring the highest marks ever on the company's automated behavioral audit.
Days after Mark Zuckerberg called Muse "built from the ground up for privacy and security," researcher Patrick Wardle found a flaw letting local apps — and a simple trick — hijack the agent's authentication token.