AI Recommendation Poisoning: How "Ask AI" Buttons Silently Alter LLM Memory
A new class of prompt injection is spreading across commercial websites. It requires no malware, no stolen credentials, and no zero-day exploit.
Red
Unknown
auto-detected
AI Intelligence Brief cached
Edit Profile
Entity Relationships
Related YARA Rules View all on YARA Rules page →
90-Day Activity
All Articles (892)
Last 90 Days
All Time
Black Hat USA 2026: One GitHub Issue Could Compromise Major AI Coding Workflows
At Black Hat USA 2026, Novee found GitHub workflow flaws in Claude Code, Gemini CLI and Codex that enabled RCE, credential theft and agent control in pipelines.
NVIDIA Group Proposes SAFE Initiative for Agentic Threat Intel Sharing
The Open Secure AI Alliance has announced plans for the Shared AI Findings Exchange (SAFE)
Hackers Abuse Cloud Startup Credits to Resell Claude and Gemini AI Access
Threat actors are exploiting free cloud trials and startup credit programs to build gray-market AI proxy services. These services resell discounted access to...
Shai-Hulud strikes again: CHAINDROP worm hits 400+ npm packages
Elastic Security Labs identified the return of Shai-Hulud. Attackers compromised the keyv maintainer and deployed CHAINDROP, a worm that uses stolen npm cred...
Shai-Hulud strikes again: CHAINDROP worm hits 400+ npm packages
Elastic Security Labs identified the return of Shai-Hulud. Attackers compromised the keyv maintainer and deployed CHAINDROP, a worm that uses stolen npm cred...
CRLF-Powered Desync Attacks: Beheading HTTP Streams
Abstract In this paper we’ll show that HTTP Header Injection is severely underestimated.
Cloudflare is the only vendor named a Visionary in 2026 SASE and SSE reports
We're honored to announce that Cloudflare is the only vendor that has been recognized as a Visionary in both the 2026 Gartner® Magic Quadrant™ for SASE Platf...
Security validation should begin where attackers begin
Modern attacks increasingly begin with the web application. Customer portals, partner platforms, APIs, external business applications, and AI-powered service...
77 malicious extensions found on Open VSX marketplace
Seventy-seven malicious extensions impersonating legitimate developer tools were discovered on the Open VSX marketplace, transmitting system and development ...
TP-Link Omada ZTP systems vulnerable to fleet-wide compromise
Fifteen new vulnerabilities were discovered in the zero-touch provisioning (ZTP) systems used by TP-Link’s Omada networking ecosystem, with researchers warni...
CAF Bank online service restored after attempted fraud and login removal
CAF Bank restored its online banking service after a prolonged outage lasting more than 10 days, which the bank attributed to attempted fraud and a subsequen...
Apple restores Telegram after child abuse material ban; Gram token recovers
Apple restored Telegram to its App Store after a brief removal due to reported child sexual abuse material.
Cisco IOS Software and IOS XE Software Extensible Messaging Client Protocol Denial of Service Vulnerability
A vulnerability in the Extensible Messaging Client Protocol (XMCP), also referred to as the External Client protocol, of Cisco IOS Software and Cisco IOS XE ...
Poison Claude Sells Discounted Claude Access While Its Operator Sees Every Customer Prompt
Cybersecurity researchers have discovered more than half-a-dozen services advertisements for illegal access to artificial intelligence (AI) models on undergr...
Google Blogger locks hundreds of blogs in malware false positive
Google has locked hundreds of Blogger websites after a false positive claimed they violated its "Malware and Similar Malicious Content" policy, with some sit...
“I’m Allowed”: Hackers Use Simple Claims to Bypass AI Guardrails
Cisco Talos found hackers using simple authorization claims to bypass AI guardrails, build DDoS attack tools, steal credentials and access live camera services.
Four Million Malware Reports Reveal a Widespread No-DNS C2 Blind Spot
A long‑running supply chain compromise of the QuickFox VPN accelerator that quietly delivered an FDMTP backdoor to carefully profiled Windows systems, exposi...
How AI-powered phishing killed blocklists for good
AI is helping attackers create disposable phishing infrastructure and rapidly evolving toolkits that blocklists cannot track fast enough. Push Security expla...
OVSwrap Open vSwitch Flaw Lets Unprivileged Linux Users Gain Root Access
A recently disclosed Linux local privilege-escalation vulnerability, tracked as CVE-2026-64531 and referred to as OVSwrap, affects the kernel’s Open vSwitch ...