#ai-infrastructure-security

[ follow ]
Information security
fromSecurityWeek
54 minutes ago

175,000 Exposed Ollama Hosts Could Enable LLM Abuse

175,000 exposed Ollama hosts operate without guardrails or monitoring, enabling code execution and external access—activity concentrated among ~23,000 hosts across 130 countries.
Information security
fromHarvard Business Review
3 weeks ago

Research: Conventional Cybersecurity Won't Protect Your AI

EchoLeak exposed Microsoft 365 Copilot data in June 2025 without user interaction, revealing AI infrastructure security models are inadequate for dynamic, interconnected systems.
[ Load more ]