GitHub Copilot Can Now Drive Desktop Apps - Off by Default, but a Saved Approval Spans CLI and App
Google Announces Gemini 4 Argon - Prices Are Published, but Fairwind Defenders Get It First
NVIDIA Launches the Open Agent Safety Platform for Containing Agents - Putting the Boundary Outside the Model
OpenAI Reports the Existence of Self-Replicating Prompt Injections That Get Copied Along Through Email and Files
OpenAI Pauses Tool-Use Training, Evaluation and Inference on Its Most Capable Models After a DNS Gap Let an Agent Out
Cursor Releases Two Bots for the Work After the PR: Rollouts and Security Review
How Claude Code's macOS Sandbox Could Be Escaped Is Now Public - Researchers Say It Was Fixed in 2.1.247
Anthropic Publishes Its September Threat Report - Distillation by Seven China-Based Labs, and Traffic Relayed to Claude Without Users Being Told
Meta Launches Personal AI Agent 'Muse' in the US — A Design That Puts Permission Decisions Outside the Model
Anthropic's Enterprise Frontier Safeguards Keeps Monitoring Logs in the Customer's Own Cloud
Cursor's Self-Hosted Machines Keep Agent Tool Execution Inside Your Own Network
HiddenLayer Raises $100M Series B for AI Security as ARR Grows 10x
OpenAI Astra Hits Critical Cyber Threshold — And May Halt Legitimate Agent Tasks
Gemini 3.8 Flash and Flash Cyber: Pricing, Benchmarks, and Who Gets Access
Claude Code 2.1.257 Makes Fable 5.1 the Default Fable Model, Adds a Containment Escape Rule
Anthropic Ships Claude Fable 5.1 and Mythos 5.1 — Same Per-Token Price, Cache Reads Cut 75%
OpenClaw 2026.8.1 Adds Masked Prompts That Keep Credentials Out of Chat
Claude Console Adds Personal Keys and Service Account Keys
Claude Code 2.1.251 Closes Several Routes Around the Permission Check
Google DeepMind Pilots Double-Blind AI Evaluations: Neither Weights Nor Test Prompts Are Shared
GitHub Code Scanning Adds a Mitigated Dismissal Reason for Alerts
GitHub Adds 8-Hour Access Tokens and Multiple Redirect URIs for OAuth Apps
Cursor Earns AIUC-1, an AI Agent Security Certification That Covers MCP
Microsoft Adds a DevSecOps Pillar to the Zero Trust Workshop — 15 Control Groups, 91 Tasks
Claude Enterprise Adds "Inference Hooks" — Every Prompt Inspected by Your Own Server First
Claude Code Makes Auto Mode the Default on August 14 — "Users Approve 97% of Permission Prompts"
GitHub Lets Enterprises Restrict Which MCP Servers Copilot Can Run - Broken Configs Fail Closed
Alibaba to Ban Employees from Using Claude Code After Hidden China-Detection Code Is Discovered