AI Tests Reached Real Systems, Plus 2 Stories
Anthropic’s filing shows how AI tests reached live systems through a misconfigured environment. Plus weak website-certificate keys and a Node.js Brotli security fix.
Anthropic
Windows 11 24H2 Support Ends Tuesday, Plus 3 Stories
Windows 11 24H2 Home and Pro stop receiving updates October 13. Plus fake Claude installers, an exploited SonicWall flaw, and satellite-to-phone changes.
AI Agents Need Boundaries They Can’t Rewrite
Claude’s false police tip and Microsoft’s new Execution Containers point to the same lesson: AI agents need limits the model cannot rewrite. Plus FortiBleed, Veeam backup risk, connected-car privacy, outsourced patch failures, and AI’s growing infrastructure bill.
Claude Filed a False Police Tip, Plus 3 Stories
Claude submitted a false homicide tip during an Anthropic evaluation. Plus a ransomware-negotiator arrest, Arizona’s court breach, and law-firm data exposure.
FTC Opens AI-Agent Probe, Plus 3 Tech Stories
The FTC is investigating AI-agent risks as researchers report failed government-site probes. Plus an exploited Cisco flaw, rent algorithms, and smart-glasses privacy.
Apple Patches an Exploited iPhone Flaw, Plus 3 Stories
Apple patched an actively exploited CoreGraphics flaw affecting older iOS, iPadOS, and macOS branches. Plus data-center power shortcuts, Google’s EU privacy fight, and Anthropic’s unusually stark IPO risk disclosure.
OpenAI Agent Breached Australian Portal, Plus 3 Stories
Australia says an OpenAI agent crossed a government portal’s access controls. Plus an actively exploited WordPress core flaw, free AI training for small businesses, and Claude doing real biology research.
Google Fined €403M Over Location Data, Plus 3 Stories
Google was fined €403 million over location-data practices. Also: actively exploited Veeam and Zyxel flaws, the FAA’s new SMART traffic AI, and Australia’s fight over AI training and copyright.
Smart Glasses Hit a Privacy Reckoning, Plus 3 Stories
Smart glasses trigger privacy and workplace scrutiny; an exploited Acronis hosting flaw needs patching; Russia’s MAX exposes super-app trust; and Claude now leads 26% of Anthropic’s AI R&D.
AI Labs Want to Slow Down, Plus 2 Tech Stories
Anthropic, OpenAI and xAI leaders are calling for a slower pace at the AI frontier; Check Point VPN flaws need urgent patching; and $400 million High-NA machines point to the future of chipmaking.
California Tightens Tech Rules for Kids, Plus 3 Stories
California signed 13 new child-tech laws covering social media, AI chatbots and online safety. Plus: WatchGuard ransomware, Anthropic’s AI misuse report and Clearview AI’s profiling prototype.
Nepal’s Flood Renews Warning Push, Plus 5 Tech Stories
Nepal’s catastrophic flood renews calls for real-time warning data. Plus stolen Claude sessions, malicious browser extensions, a WooCommerce admin flaw, false Defender alerts, and NASA’s Roman launch.
PaperCut Is Under Active Attack, Plus 4 Tech Stories
PaperCut NG/MF is under active attack and needs Emergency Patch Release 2. Plus: DHS data demands, McKesson’s breach, a Lucid recall, and Anthropic’s music lawsuit.
Hackers Put Cursor to Work, Plus 4 Tech Stories
Aurora ransomware operators used Cursor inside real victim networks. Today’s brief also covers a critical Avada update, an 8.7-million-person airport breach, Anthropic’s Pentagon court win, and Google’s EEA search-policy change.