Baseten details security hardening approach for multi-cloud clusters
Understand this story
Baseten published a blog post describing how it maintains a unified security standard across more than 20 clouds and over 100 clusters (including more than ten neoclouds). The company says its threat model is based on four trust boundaries: the infrastructure layer, operational partners, workloads, and internal access, and that each cluster is hardened before it receives traffic. At the customer workload level, customer builds run in microVMs, each with its own kernel; tenants are isolated from one another by network policies, cross-cluster traffic uses mTLS, and credentials cover at most a single cluster. Baseten also says engineers have no long-term administrative access to production clusters, and that logs for all access paths are exported outside the cluster. The above practices and results all come from Baseten's own account; the blog post provides no third-party verification or specific incident data.
AI overview based on source reports · Updated 2 hr ago
Coverage timeline
Follow the coverage to explore different perspectives.
- Baseten BlogHow we secure 100+ clusters across 20+ clouds
Baseten explains how it applies a consistent security standard across 20+ clouds and 100+ clusters, including over a dozen neoclouds, by starting from a threat model with four trust boundaries: the infrastructure layer, operating partners, workloads, and internal access. It describes controls such as closing public ports, allowlisting the Kubernetes API, per-cluster read-only deploy keys, per-cluster certificate authorities, hourly credential rotation, microVM-isolated builds, and centralized logging with daily external perimeter scans. The post ends with a checklist for auditing a fleet or evaluating a provider's security standards.
Activity for this story
Not enough continuous observations to show a trend yet.