# Baseten — How we secure 100+ clusters across 20+ clouds

- Company: Baseten (baseten.co)
- Announced: 2026-11-20T17:10:28+00:00
- Category: not stated
- Coverage: not counted
- Announcement: no
- Group: routine
- Source: https://www.baseten.co/blog/how-we-secure-100-clusters-across-25-clouds
- Record: https://forck.live/items/39317-how-we-secure-100-clusters-across-20-clouds
- Subject: Baseten

Baseten has published a walkthrough of how it holds one security standard across a fleet of 20+ clouds and 100+ clusters, including more than a dozen neoclouds it uses for global scale. The company says it treats GPUs as fungible across providers, so every cluster is evaluated and hardened before it serves traffic, because neoclouds are often less hardened and less uniform than hyperscalers. Baseten describes a threat model with four trust boundaries: the infrastructure layer, operating partners, workloads, and internal access. At handover it closes every public port except those serving traffic, puts the Kubernetes API behind an allowlist, and sets SSH to key-only access with regularly rotated keys, with on-demand SSH behind multi-factor authentication and multi-party approval. Customer builds run in microVMs, tenants are isolated on the network, and RDMA transfers pass through a node agent that authorizes each one. Baseten says nobody there has standing administrative access to production clusters, and that it scopes credentials to one cluster at most so a compromised cluster cannot reach another. The post ends with a checklist readers can apply to their own fleet or use when evaluating a provider's security standards.

## Evidence

Verbatim from https://www.baseten.co/blog/how-we-secure-100-clusters-across-25-clouds:

> Because we treat compute as fungible across many GPU providers, we have to take steps to ensure our security standard holds on every cluster we run on. This post covers how we do it, including a checklist at the end you can apply to your own fleet, or when evaluating a provider’s security standards.
> 
> Running inference for other companies requires a high bar of trust, and we have to continuously earn that trust when it comes to serving our customers’ models, processing their data, and offering the products they build on both. We're proud to partner with teams like Cognition, Coinbase, Cursor, Datadog, HubSpot, Ramp, and many more, and we put controls in place to protect their models and data wherever they run.
> 
> This post covers how we hold our security standard across our entire fleet—20+ clouds and 100+ clusters—including over a dozen neoclouds that help us unlock global scale (and run workloads active-active) to reliably meet any level of customer demand. For an introduction to this security series and a look at how we’ve adapted our defenses to agent-speed attacks, read our overview blog here .
> 
> Why every cluster needs its own hardening

---

Record: https://forck.live/items/39317-how-we-secure-100-clusters-across-20-clouds
Catalogue: https://forck.live/llms.txt
Current issue: https://forck.live/feed.md
