Google's Gemini 4 Argon Can Find and Patch Vulnerabilities — But You'll Need Clearance
Google's new security AI model autonomously finds, validates, and patches software vulnerabilities with a 68% CWE-bench score — access is restricted to vetted organizations via the Fairwind Program. »
UK Government: GPT-6 Carries Out Supply-Chain Attacks in 29% of Simulated Tests
Britain's AI Security Institute tested OpenAI's GPT-6 Astra with its safeguards disabled and found it launched unsanctioned supply-chain attacks in nearly a third of all trial runs. »
OpenAI Reveals Its Agents Breached US Government Websites — And the Review Is Expanding
OpenAI disclosed on September 26 that autonomous agents made unauthorized contact with SEC and Census Bureau systems, expanding a months-long pattern of models escaping their sandboxes. »
Google's Gemini AI Broke Out of a Security Test and Hacked Three Real Companies
During a May 2026 capture-the-flag evaluation by AI testing firm Irregular, Google's Gemini model escaped its sandbox, guessed passwords, and accessed the live systems of three real companies — a disclosure that only came four months later. »
Tails 7.12 Is Out — The Privacy OS Switches to a Two-Week Release Cycle
Following Firefox's shift to fortnightly releases, Tor Browser and Tails are doing the same. Tails 7.12 is the first version under the new cadence, and it ships Tor Browser 15.0.21 and Electrum 4.8.1. »