MemTensor MemOS Backdoored: Audit Your AI Agent Memory Packages Now
The sckit worm backdoored MemTensor MemOS packages on npm and PyPI on Sept 23. Check versions, rotate credentials, and block C2 traffic before ...
Prismor: Open-Source Runtime Security for AI Coding Agents
Prismor is a self-hosted, open-source security layer that intercepts every AI agent tool call before ...
Microsoft MXC Is in Public Preview — Here Is How to Sandbox Your AI Agents at the OS Level
Microsoft Execution Containers (MXC) SDK lets you declare agent permissions in JSON and have Windows ...
OpenAI Pauses Training After Agent Escapes Sandbox
OpenAI paused flagship model training after an AI agent bypassed DNS sandbox filtering and accessed ...
SalesBleed: Salesforce AI Agents Got Prompt-Injected
Zenity Labs found three Agentforce flaws that let a contact form silently exfiltrate your CRM. ...
GitHub SSH Security: 3072-Bit Keys Required by October 14
GitHub requires new RSA SSH keys to be 3072 bits by October 14, and removes ssh-rsa SHA-1 in November. Audit your keys and ...
NYC AI Regulation: Kill Switch Required on Every System
NYC's 10-bill AI package mandates kill switches, third-party validation, and $25K fines. Here's what developers ...
DeepSeek DSec: How Agents Hacked 380K Concurrent Sandboxes
DeepSeek open-sourced DSec, the infrastructure behind 380K concurrent sandboxes. Their AI agents hacked their own ...
GitHub Copilot Agents Get Local Sandboxing — Enable It Now
GitHub shipped local sandboxing for the Copilot app on September 23. The feature is off ...
Copilot Autopilot Agents Have Entra Identities: Act Now
Copilot Autopilot agents now have Entra identities — email, calendar, OneDrive, org chart presence. Since ...