Fetching from the wire…
01
02
03
04
05
06
07
08
Infra2026-09-22 · source-backed
The architecture executes untrusted agent-generated code on Bedrock AgentCore Code Interpreter in VPC mode with no internet or NAT gateway, port 443 only, prefix-list routing confined to VPC endpoints, and a three-tier Route 53 DNS firewall whose catch-all rule returns NODATA for anything not explicitly allowed. Isolation is per-job STS credentials scoped to one tenant's S3 prefix rather than static per-tenant roles, with VPC endpoint policies enforcing bucket access at the network layer independent of IAM. 600+ sessions daily, no security incidents since the April 2026 deployment, and integration tests that actively simulate DNS tunneling to catch config drift. (AWS)
Each link below shares sources, entities, or timing with this story.
AWS shipped Policy in Amazon Bedrock AgentCore to general availability across 13 regions. The breakthrough: security teams can write agent-to-tool access rules in plain English, which auto-convert to Cedar policies with automated reasoning that catches overly permissive or uns...
Each tenant gets a dedicated Bedrock AgentCore runtime, every user session runs in its own microVM that's terminated and memory-sanitized on completion, paired with Knowledge Bases metadata filtering, Guardrails on responses, VPC Lattice for private connectivity and per-tenant...
Loom (github.com/awslabs/loom) is a reference implementation: "paved path" blueprints baking in least-privilege IAM, abstracted configuration, and strict deployment guardrails for agents built on Strands Agents running on Bedrock AgentCore Runtime. The repo ships the specs use...
Announced September 8, running on the Bedrock inference engine. AWS An availability item, and it matters if you need Astra-class reasoning inside an AWS account boundary with existing IAM and VPC controls. ---
The full family, Sol, Terra, and Luna, is generally available on Amazon Bedrock with IAM and VPC controls (LLM Boss, AWS). Sol targets coding, biology, and cybersecurity agentic work. Terra runs everyday tasks at about half GPT-5.5's cost, and Luna optimizes for speed. The thr...
AWS shipped per-user, per-target rate limiting August 6 covering MCP targets, inference targets and HTTP passthrough across three metrics: requests (RPS/RPM), tokens (TPM, inference only), connections (CPS). Limits scope by JWT claims ($.context.jwt.sub, .role, .azp) or IAM id...
MindPattern daily
One email a day at 7 AM. Sources and a take on every story. Unsubscribe anytime.