Concept  ·  Glossary

AI inference serving infrastructure

The servers and software (like vLLM, SGLang, or GPUStack) that host a trained AI model and run it for users on demand. If an attacker compromises this layer they can run their own code, read other tenants' data, or shut the service down. It is the AI equivalent of a data centre — the place where models actually execute.
This is where real models run, so intrusions here are often the most damaging and hardest to detect, and these platforms are now appearing on national 'known exploited' lists.
CISA Known Exploited Vulnerabilities (KEV) CatalogCLOUD SECURITY ALLIANCE GPUStack CVE advisory
Track this in the live feed See how this plays out in real AI security and governance developments.
Open the feed →