#ai-security
325 stories taggedai-security · page 17 of 22.
Agentic AI Runs on Context. Feed It the Wrong Kind and Decisions Go Sideways Fast.
The core vulnerability in agentic AI systems isn't the model — it's the context window. Bad inputs, machine-speed outputs.
Anthropic's AI Model Found Vulnerabilities in Classified U.S. Government Systems
An unnamed U.S. official says Anthropic's Mythos model identified security flaws in sensitive government infrastructure during a joint exercise with intelligence agencies.
Fake Agent Skill Slips Past Every Scanner, Lands on 26,000 AI Agents
A red-team experiment by AIR pushed a booby-trapped skill through a popular marketplace and an Instagram ad. The skill marketplaces' security scanners shrugged.
When the Trigger Pulls Itself: Agentic AI and the End of the Human-in-the-Loop
Every weapon in history extended a human decision. Agentic systems are the first that try to replace it — and the security implications are not theoretical.
OpenAI Hands GPT-5.5-Cyber to 'Trusted Defenders' Under Daybreak
The model is pitched at deep codebase analysis and vuln patching. The interesting part is who gets access — and what shows up in the post-mortem when they don't.
DifyTap: Four Unauthenticated Bugs in Dify Expose Cross-Tenant AI Conversations
Researchers at Zafran say a chain of flaws in the popular agentic workflow platform let attackers read other tenants' chats without logging in.
AWS Continuum Wants to Close the Gap Between AI-Generated Code and AI-Fixed Vulnerabilities
Amazon's new agentic security service promises continuous discovery, triage, and remediation. In practice, it's a bet that the same AI acceleration creating your backlog can also drain it.
When Legacy Infrastructure Becomes the Soft Underbelly of Your AI Agent Stack
Governance frameworks like NIST AI RMF and the EU AI Act assume the pipes under the model are secure. They often aren't.
AutoJack: When the AI Browser Becomes the Initial Access Broker
Microsoft researchers describe an exploit chain that turns an agentic browser into a one-click path from web page to host process execution.
The SOC Triangle Was Always a Lie We Accepted. AI Is Changing the Math.
Security operations have run on a structural compromise for decades — quality, consistency, or cost: pick two. That constraint is finally starting to bend.
AutoJack Exploit in Web-Enabled AI Agents: Bypassing Localhost Security
Microsoft uncovers RCE vulnerability in AutoGen Studio through local AI agent misuse.
Tool Sprawl Meets Agentic AI: Why SOCs Are Rethinking the Triage Stack
Forty tools, forty-three day dwell times. Vendors are pitching agentic AI as the fix. Analysts have questions.
SearchLeak Shows How a Single Crafted URL Can Drain Your M365 Tenant
Varonis researchers chained three weaknesses in Copilot Enterprise Search into a full data-exfiltration path. Microsoft patched it. The attack class isn't going anywhere.
Browser Add-Ons, AI Chat Links and In-Memory macOS Attacks: A Week the Internet Worked As Designed
Shady extensions, weaponised Claude conversations, fileless macOS intrusions and cloud agents turned into shells dominated the criminal feeds this week.
Mastra npm Namespace Hit: 145 Packages Tampered After Contributor Account Hijack
Researchers tracking the 'easy-day-js' supply chain incident say a single compromised maintainer account was sufficient to push malicious versions across the @mastra/* registry footprint.