AIKIDO-2026-63881
@openai/agents-openai is vulnerable to Improper Authorization
59
Medium Risk
This Affects:
@openai/agents-openaiTL;DR
The Responses model conversion maps the hosted MCP require_approval policy without validating it before sending it to the API. An undefined or malformed policy and overlapping always/never tool lists are propagated as-is, and undefined is treated as never requiring approval. This can silently disable the intended human-in-the-loop approval gate so hosted MCP tool calls execute without approval. The fix routes the policy through a shared validator that normalizes it and rejects invalid configurations.
Who does this affect?
You are affected if you are using a version that falls within the vulnerable range and configure hosted MCP tools with require_approval approval policies.
Background info
@openai/agents-openai is vulnerable to Improper Authorization in versions 0.0.7 - 0.10.0.
How to fix this
Upgrade the @openai/agents-openai library to the patch version.
Are You Affected?
Connect your repositories to instantly see whether vulnerable or malicious packages exist in your codebase.
Free. No credit card required.
I consent to receiving marketing communications based on Aikido’s Privacy Policy.
SOC 2Compliant
ISO 27001CompliantISO 42001Compliant
FedRAMPImplementing© All Intel data is openly available and commercially licensed.