CoreBreak Flaws Let Attackers Invoke AWS, Google, Vercel Tools
August 6, 2026
Stealth researchers disclosed CoreBreak flaws in AWS Bedrock, Google ADK, and Vercel harnesses that let attackers invoke agent tools without
News
CoreBreak Flaws Let Attackers Invoke AWS, Google, Vercel Tools
Stealth researchers disclosed CoreBreak flaws in AWS Bedrock, Google ADK, and Vercel harnesses that let attackers invoke agent tools without the model.
ClickFix Malware Gate Fingerprints macOS Users Before Lures
Microsoft detailed a ClickFix campaign spanning 250 domains that fingerprints macOS visitors server-side before deciding whether to show an infostealer lure.
Cisco Patches Critical SD-WAN, IOS XE, and FMC Flaws
Cisco patched two dozen flaws including critical Catalyst SD-WAN and IOS XE command-injection bugs plus an FMC authentication bypass in a new advisory release.
Open VSX Purges 77 Evil-Twin Extensions Stealing Developer Data
Open VSX removed 77 malicious evil-twin extensions impersonating developer tools and exfiltrating machine, Git, and CI/CD data to one attacker domain.
ChainDrop npm Worm Poisons 440 Packages, Steals Cloud Credentials
The ChainDrop npm worm, a new Shai-Hulud variant, poisoned over 440 registry packages and uses stolen tokens to republish malware and reach cloud credentials.
CISA Adds Exploited Langflow and Tomcat Flaws to KEV Catalog
CISA added actively exploited Langflow and Apache Tomcat vulnerabilities to the KEV catalog, linking the Tomcat flaw to an AI-enabled Chinese hacking campaign.
QuickFox VPN Supply-Chain Attack Delivers FDMTP Backdoor
Fortinet disclosed a long-running supply-chain attack on QuickFox VPN that delivers the undocumented FDMTP backdoor through a trojanized Windows installer.
SMOKE#SCREEN Deploys ScreenConnect via Fake Adobe, Zoom Lures
Securonix details the SMOKE#SCREEN campaign, which uses fake Adobe and Zoom update lures to stealthily install ConnectWise ScreenConnect for persistent access.
Claude Mythos 5 Tried to Backdoor a Project in UK AI Security Test
A Claude Mythos 5 agent spent 34 hours trying to merge malware into an open-source project during a UK AI Security Institute evaluation, then covered ...
Google Deletes Three ADK AI Workflows After Prompt-Injection Attack
Google removed three ADK AI workflows after Pillar Security showed a GitHub issue could prompt-inject a triage agent into launching a privileged agent.
ClickFix Malware Gate Fingerprints macOS Users Before Lures
August 6, 2026
Microsoft detailed a ClickFix campaign spanning 250 domains that fingerprints macOS visitors server-side before deciding whether to show an infostealer
Cisco Patches Critical SD-WAN, IOS XE, and FMC Flaws
August 6, 2026
Cisco patched two dozen flaws including critical Catalyst SD-WAN and IOS XE command-injection bugs plus an FMC authentication bypass in
Open VSX Purges 77 Evil-Twin Extensions Stealing Developer Data
August 5, 2026
Open VSX removed 77 malicious evil-twin extensions impersonating developer tools and exfiltrating machine, Git, and CI/CD data to one attacker
ChainDrop npm Worm Poisons 440 Packages, Steals Cloud Credentials
August 5, 2026
The ChainDrop npm worm, a new Shai-Hulud variant, poisoned over 440 registry packages and uses stolen tokens to republish malware
CISA Adds Exploited Langflow and Tomcat Flaws to KEV Catalog
August 5, 2026
CISA added actively exploited Langflow and Apache Tomcat vulnerabilities to the KEV catalog, linking the Tomcat flaw to an AI-enabled
QuickFox VPN Supply-Chain Attack Delivers FDMTP Backdoor
August 5, 2026
Fortinet disclosed a long-running supply-chain attack on QuickFox VPN that delivers the undocumented FDMTP backdoor through a trojanized Windows installer.
SMOKE#SCREEN Deploys ScreenConnect via Fake Adobe, Zoom Lures
August 5, 2026
Securonix details the SMOKE#SCREEN campaign, which uses fake Adobe and Zoom update lures to stealthily install ConnectWise ScreenConnect for persistent
Claude Mythos 5 Tried to Backdoor a Project in UK AI Security Test
August 5, 2026
A Claude Mythos 5 agent spent 34 hours trying to merge malware into an open-source project during a UK AI
Google Deletes Three ADK AI Workflows After Prompt-Injection Attack
August 5, 2026
Google removed three ADK AI workflows after Pillar Security showed a GitHub issue could prompt-inject a triage agent into launching
Trending
Daily Briefing Newsletter
Subscribe to the Daily Security Review Email Briefing to stay informed on the latest threats, trends, and technology, along with insightful columns from industry experts.