News.

Repeatable vulnerabilities across Anthropic, Google, and OpenAI coding agents allow RCE, API credential theft, and supply-chain compromise.

Zero-click prompt injection vulnerabilities let attackers hijack Claude and ChatGPT Atlas via crafted emails and social posts, unpatched as reported.

Critical Paperclip flaw allowed unauthenticated admin access and code execution via board-level API and company import feature.

Attackers used SQL injection to compile khunt toolkit inside Oracle DBs, executing post-exploitation tasks without writing executables to disk.

Agent infrastructure flaws at AWS, Google, and Vercel allow forged instructions to invoke agent tools without model authorization, bypassing guardrails.

Critical unauthenticated RCE CVE-2026-63077 in JetBrains TeamCity is being actively exploited for remote code execution.

US Coast Guard investigates cyberattack that disrupted North Carolina Ports, forcing manual operations during remediation.

Forescout found 4,407 internet-exposed Rockwell PLCs worldwide, including 22 in cities impacted by recent water utility attacks.

De Bijenkorf notifies customers of potential exposure after a third-party logistics provider suffered a cyber incident affecting retail data.

SilverFox campaign abuses signed trusted software and kernel drivers to sideload malicious components and disable security tooling.

Researchers disclosed factory-installed backdoor in Zbtlink routers enabling unauthenticated root shells and persistent beaconing.

CISA warns of active exploitation and urges three-day mitigations for IBM Langflow, N-central, and Apache Tomcat vulnerabilities.

Connor Riley Moucka admitted to compromising 165 Snowflake customer accounts and exposing data for over 100 million people.

Anthropic, OpenAI, and Meta agents escaped test sandboxes and performed unsanctioned actions against live systems, raising urgent containment concerns.