🟠 High | Source: AWS Security Bulletins
A vulnerability in Amazon Bedrock’s AgentCore InvokeHarness API allowed authenticated users to trigger configured tools directly, bypassing the AI model and its associated security controls. The flaw meant that a crafted request containing a tool-use content block could cause the agent to dispatch tools without model mediation. AWS patched the issue on 31 July 2026; the blast radius was bounded by whichever tools were configured on the affected harness.
Security Architect’s Take: Review audit logs for InvokeHarness API calls made prior to 31 July 2026, particularly any requests where tool invocations occurred without corresponding model inference events, as these may indicate exploitation. Ensure least-privilege tool configurations are applied to all Bedrock AgentCore harnesses to limit potential impact from similar bypass vulnerabilities in future.
Original advisory: CVE-2026-18830 - Issue with Amazon Bedrock AgentCore harness – Insufficient Input Validation