Anthropic and OpenAI Both Disclose Rogue Agent Incidents Involving Real-World Victims
Two leading AI labs have now admitted their models escaped sandboxed test environments and attacked…
Month: July 2026
Generic TV Streaming Sticks Linked to Ad Fraud Network Generating Estimated $50,000 Per Day
A Bitsight researcher's analysis of an expired domain reveals how H96 streaming devices silently conduct…
CISA Adds Cisco Firewall Management Centre Hard-coded Password Flaw to Known Exploited Vulnerabilities Catalogue
Active exploitation of a hard-coded credential vulnerability in Cisco Secure Firewall Management Center prompts a…
Credential stuffing campaign hits 30 SonicWall customers in under 48 hours
A fast-moving, opportunistic attack wave compromised 92 SonicWall accounts across 30 organisations in 41 hours,…
OpenAI Models Exploited JFrog Artifactory Zero-Days to Escape Sandbox and Breach Hugging Face
Eight previously unknown vulnerabilities in Artifactory's package registry cache proxy allowed AI models under evaluation…
AI Finds More Vulnerabilities Than Attackers Currently Exploit, Research Suggests
A VulnCheck analysis finds that just one per cent of vulnerabilities discovered with AI assistance…
OpenAI Discloses Rogue Agent Accessed Multiple External Services Using Exposed Credentials
A new OpenAI disclosure reveals its AI agent autonomously used leaked logins to breach at…
MikroTik RouterOS Authentication Flaw Has No Patch — Mitigations Are Your Only Option
A CISA-published advisory details a brute-force vulnerability affecting all versions of MikroTik RouterOS and Cloud…
OpenAI Agent’s Accidental Attack on Hugging Face Offers a Detailed Blueprint of Machine-Speed Intrusion
A five-day autonomous campaign exploiting a JFrog Artifactory zero-day, Kubernetes token theft, and a self-provisioned…
NVIDIA and 36 Partners Form Open Secure AI Alliance, Release NOOA Security Framework
A broad coalition of cloud, security, and AI vendors is pooling resources to build open…