When Anthropic announced Project Glasswing, it felt like a turning point in how we think about cybersecurity. Rather than another incremental tool, Glasswing pools one of the most capable frontier language models—Claude Mythos Preview—with an unusual, urgent mission: give the organizations that run the internet and financial systems a head start against AI-enabled attackers. The initiative reads like a playbook
Category: Cybersecurity
Zero-Day, APT, Exfiltration, Lateral-Movement, Privilege-Escalation, Botnet, Rootkit, Backdoor, Keylogger, Smishing, Vishing, Spear-Phishing, Social-Engineering, MITM, SQL-Injection, XSS, CSRF, Path-Traversal, Buffer-Overflow, Honeypot, CVE, CVSS, Red-Team, Blue-Team, Threat-Hunting, Malware-Analysis, MITRE-ATT&CK, Insider-Threat, Jailbreak, Shellcode, Exploit-Kit, LFI, RFI, Obfuscation, Payload, security advisory, vulnerability disclosure, CWE, OWASP, cybersecurity news, threat intelligence, SOC, SIEM, cryptotheft, evasion, CVE Security
ActiveMQ broker RCE tied to CVE-2026-34197: what admins need to know
A long-standing flaw in Apache ActiveMQ has resurfaced as a serious concern for administrators. The issue—listed on CISA’s Known Exploited Vulnerabilities (KEV) list under entry 46604—enables unauthenticated remote command execution via the broker port. Although CVE-2026-34197 is not yet reported as being widely exploited in the wild, researchers examining broker logs say there are clear indicators that attackers have attempted
Anthropic Withholds Mythos Preview: Too Potent a Cyber Threat to Release
Anthropic’s decision to withhold the Claude Mythos Preview has punctured the usual celebratory arc of model announcements. Rather than rushing to commercialize another frontier AI, the company says Mythos demonstrated capabilities that could be exploited to find and chain high-severity vulnerabilities in widely used systems—so serious that Anthropic is choosing limited, defensive deployment over general release. A startling discovery in
OpenAI Codex Command-Injection Flaw: How GitHub Tokens Were Exposed and What Teams Must Do Now
The rise of AI coding assistants has simplified developer workflows, but a recent discovery shows those conveniences can carry serious risk. Researchers at BeyondTrust found a critical command-injection vulnerability in OpenAI Codex that could be exploited to steal GitHub access tokens. The flaw demonstrates how an overlooked parsing detail — a branch name passed into a container setup script —
Critical RCE in Ninja Forms File Upload Exposes ~50,000 WordPress Sites
A recently disclosed vulnerability in the popular Ninja Forms “File Upload” addon has placed roughly 50,000 WordPress sites at risk of full takeover. Tracked as CVE-2026-0740 and carrying a CVSS score of 9.8, the flaw allows unauthenticated arbitrary file uploads — a straightforward path to remote code execution (RCE) for attackers. Site owners who rely on the affected plugin must
Researcher Publishes Windows Defender 0-Day ‘BlueHammer’ LPE Proof‑of‑Concept
A security researcher using the handle Chaotic Eclipse has publicly released a working proof‑of‑concept for a Windows zero‑day local privilege escalation (LPE) exploit called “BlueHammer.” The disclosure, accompanied by full source code on GitHub, was confirmed as functional by vulnerability researcher Will Dormann and demonstrates that a low‑privileged local user can escalate to NT AUTHORITYSYSTEM on affected machines. The release