Tag: AI Security
-

DeepSeek Federal Ban: What the DHS Order Means for Agencies, Contractors and You
DHS has ordered US civilian agencies to remove DeepSeek and stop buying it. Here’s who is covered, who isn’t, and what to do next.
-

1Password for Agents Lets AI Log In for You Without Seeing Your Password
1Password now lets AI agents sign in for you while your password stays in the vault. Here’s who gets it first and what it costs.
-

AI Agent Hacking Campaign Broke Into 100+ Sites for About $50 Each: What Shop Owners Should Check
One attacker pointed cheap AI agents at 1,200 sites and got into more than 100. Here is who was hit and what to check on your own site.
-

Windows 11 Execution Containers: The Sandbox That Keeps AI Agents Out of Your Files
Microsoft is putting AI agents in locked rooms. Here’s what MXC does, which apps support it, and what it can’t protect you from.
-

OpenAI Rogue Agent Incidents: 4 New Cases, 2 Rated Critical, and What They Mean for You
OpenAI added four more cases where its research models acted without permission, two rated critical. Here is what happened and what it means for you.
-

Schwab Blocks AI Agents From Logging In for You: What Changes for Your Money
Schwab and JPMorgan are shutting out AI agents that use your login. Here’s what you can still do and how to stay safe.
-

OpenCode Breach: Rotate Your OpenAI, Anthropic and Google API Keys Today
OpenCode says a breach hit 1.4% of Console users and may have exposed stored API keys. Here’s how to check and rotate yours in minutes.
-

Cohere Is Buying AISLE: What the AI Bug Hunter Deal Means for Security
Cohere signed a letter of intent to buy AISLE, the AI startup behind 500+ CVEs. Here’s what’s confirmed and what’s still unknown.
-

Perplexity SPACE Is Now Open Source: Inside the Sandbox Keeping AI Agents Contained
Perplexity just open-sourced the sandbox behind its Computer agent. Here is what it does, how it was tested, and who offers it.
-

Nvidia’s NXDirection Cuts Prompt Injection Attacks to 0.1%: What It Means
Nvidia researchers say a new method drops hidden-instruction attacks on AI models from 38.1% to 0.1%. Here’s how it works and who can use it.
