When AI Agents Escape Sandboxes, Old Security Rules Apply

Dark Reading feed covering active and emerging cybersecurity threats, including AI-agent sandbox escapes and autonomous attacks, cloud and identity takeover vulnerabilities, zero-days, ransomware, phishing, malware, data exposure, and weaknesses in passkey and AI-security implementations. Several items describe actively exploited or high-impact vulnerabilities, notably WordPress remote takeover flaws, Zimbra zero-day exploitation, SonicWall SMA zero-days, Azure cross-tenant identity takeover, and Active Directory certificate privilege escalation.