How Benchling secured multi-tenant AI agents with Amazon Bedrock AgentCore
Not a pilot. Benchling deployed AI agents running untrusted code across thousands of life-sciences tenants—here's how they locked it down.
Why it matters
Multi-tenant agent security is becoming a solved problem. Benchling's defense-in-depth architecture (sandboxing + DNS firewall + VPC policies) shows how to run autonomous code generation safely at scale—a blueprint for enterprises deploying agents across customer data.
The key facts
11 to knowBenchling deployed AI agents across thousands of life-sciences tenants
Used Amazon Bedrock AgentCore Code Interpreter in VPC mode for sandboxing
Combined Route 53 Resolver DNS Firewall with VPC endpoint policies to block data exfiltration
Architecture addresses untrusted, agent-generated code execution
Defense-in-depth approach: sandboxing + DNS controls + network policies
Benchling deployed AI agents across thousands of tenants
Used Amazon Bedrock AgentCore Code Interpreter in VPC mode
Implemented defense-in-depth: Route 53 DNS Firewall + VPC endpoint policies
Focused on blocking data exfiltration, including DNS-based attacks
Life sciences / scientific code execution as use case
Multi-tenant isolation and untrusted code execution as core problem
Go to the source
AWS Machine Learning Blogaws.amazon.com
Publisher excerpt: Learn how Benchling built a defense-in-depth security architecture to run untrusted, AI agent-generated scientific code across thousands of life sciences tenants using Amazon Bedrock AgentCore Code Interpreter in VPC mode, combined with Amazon Route 53 Resolver DNS Firewall and VPC endpoint…