Control which domains your AI agents can access
New AWS Network Firewall integration lets enterprises restrict AI agents to approved websites only.
Amazon has introduced a critical security enhancement for its Bedrock AgentCore platform, enabling enterprises to control exactly which internet domains their AI agents can access. The solution integrates AWS Network Firewall to perform TLS Server Name Indication (SNI) inspection, allowing administrators to create domain allowlists, block entire categories like social media or gambling, and log all connection attempts for audit trails. This architecture deploys AgentCore Browser tools within a private Amazon VPC subnet, routing all outbound traffic through the firewall for enforcement before reaching the internet via a NAT Gateway.
This capability directly addresses the top security concerns of regulated industries and multi-tenant SaaS providers deploying AI agents. Security teams can now prevent agents from being tricked into navigating malicious sites via prompt injection attacks by restricting access to a pre-approved list of domains. The managed rules also help block access to known botnets and malware domains. For multi-tenant environments, this enables per-customer network policies—Customer A's agents can access different domains than Customer B's—fulfilling complex compliance requirements around data sovereignty and regional restrictions.
- Uses AWS Network Firewall SNI inspection to filter domains AI agents can access, enabling allowlists and category blocks
- Deploys AgentCore Browser in private VPC subnets with no direct internet, routing all traffic through firewall controls
- Provides managed rules to block botnets/malware and enables per-customer policies for multi-tenant SaaS providers
Why It Matters
Enables regulated enterprises to deploy AI agents for web tasks without risking data leaks or compliance violations from uncontrolled browsing.