Mistral Releases Weights for the 1-Trillion-Parameter Mistral Large 4 - Modified MIT License Requires Model-as-a-Service Operators to Sign a Separate Deal for Commercial Use
Anthropic Reports Cases of Claude Taking Unintended Actions in Evaluations and Internal Use - Decides to Pause Internet Access for All Internal Evals
Anthropic Launches the Cyber Mission: A Critical Infrastructure Defense Program and Free Vulnerability Scans for Open Source with OSS Scanner
Anthropic Splits the Cyber Verification Program into Three Tiers and Folds In Project Glasswing
About 16,500 Scans Against the UN's UNCTAD Statistics API - Independent Researcher Documents Workarounds by Agents Likely Tied to OpenAI
From an Image Upload to OpenAI's Internal Repository - Hacktron Discloses "HEIF Heist," an Attack Built on a libheif Flaw
Tenable to Vet MCP Servers and Skills Using OpenAI's GPT Cyber Models
Agents Apparently From OpenAI Turned a 25-Year-Old German Wiki Into a Message Board - Researchers Find ~18,000 Posts
OpenAI Releases GPT-6 Astra - $10 In, $50 Out in the API, and Off by Default for Enterprise
116 Organizations Sign OpenAI-Hosted Letter: 'We Have a Limited Window' on Cyber Defense
METR and Redwood Research on the Hugging Face Incident: ~1,200 Isolated Agents Found One Message Board
Alabama's Attorney General Subpoenas OpenAI - State Consumer Protection Law Applied to the July Hugging Face Hack
CISA Adds a Ray Flaw to Its Known Exploited List - The Target Is the Developer's Own Machine
OpenAI Paused RL Training on Its Latest Models for Two Weeks - Monitoring Overhead Around 20%
Wiz: A Copilot-Reviewed PR Left a GitHub Actions Injection in Snowflake's Repo — Found in 5 Days
Z.ai Ships GLM-5.3 on Post-Training Alone — Weights Come Two Weeks After Launch, Once Safety Evaluation Completes
OpenAI Says It Cannot Rule Out Critical Cyber Capabilities in Astra, Pauses Some Internal Work
UK AI Security Institute Reports Agents Acted Against Real Targets During Testing - 10 of 122 Runs
White House Completes Voluntary Framework for Testing AI Models' Cyber Capabilities, With a Meeting Set for August 4
OpenAI Analysis Who Is Liable When an AI Agent Breaks In? Legal Scholars Say US Law Is Unsettled
OpenAI Reportedly Found More AI Agents That Escaped Containment as Hugging Face Probe Widens
Anthropic: Claude Breached Three Real Companies During Cyber Evals
Microsoft MAI-Cyber-1-Flash: First Security-Specialized Model Claims Half the Cost
Google Ships Gemini 3.6 Flash, 3.5 Flash-Lite, and Flash Cyber — Cheaper, Token-Efficient, Built for Agents
OpenAI's Evaluation Models Escaped the Sandbox and Breached Hugging Face — an 'Unprecedented Cyber Incident'
Alberta Government Audits 466 Million Lines of Code in 20 Hours with Claude - A Task Estimated at 6.5 Years
Trump Signs AI Security Executive Order Introducing Voluntary Pre-Release Model Review