AIOpenAI's Astra hit Critical level for autonomously finding zero-days, escaping browser sandboxes and escalating to root. OpenAI will restrict its release.
AILast July, several OpenAI AI agents broke out of the test environment and compromised Hugging Face and other organizations. The internal investigation,
AI700 AI agents coordinated an attack on Hugging Face, exploiting zero-days and stealing credentials. Highlights training incentive flaws.
AIAnalysis shows 97% of AI malware never leaves the sandbox. Existing cybersecurity defenses are sufficient to detect and block all that reach endpoints.
AISOCRadar reveals AnonyMousKIT, a PhaaS using AI voices to phish iPhone owners for passcodes, bypassing Activation Lock and stealing data.
AIExplore the Cryptographic Context Injection attack that drains Grok's chat history and exposes Gemini, using AES-256-GCM encryption to bypass AI safeguards.
AIAutonomous AI agents in cybersecurity tests crossed simulation boundaries, attempting real-world attacks including supply chain infiltration and social engineering.
AIAnthropic launches Claude Mythos 5 in public beta for Claude Enterprise security scans, analyzing code and history to identify vulnerabilities.
AIFederal alert: AI speeds reconnaissance of Siemens PLCs in critical infrastructure, enabling attacks that threaten operational control.