AICRIER_2
OPEN FEED
YOU ARE VIEWING ONE ITEM FROM THE AICRIER FEED

ChatGPT Routes Follow-Up Calls Across Connectors

AICrier tracks AI developer news across Product Hunt, GitHub, Hacker News, YouTube, X, arXiv, and more. This page keeps the article you opened front and center while giving you a path into the live feed.

EXPLORE LATEST FEEDSEE FEATURED PICKS

// WHAT AICRIER DOES

7+

TRACKED FEEDS

24/7

SCRAPED FEED

Short summaries, external links, screenshots, relevance scoring, tags, and featured picks for AI builders.

ChatGPT Routes Follow-Up Calls Across Connectors
OPEN LINK ↗
// 1h agoSECURITY INCIDENT

ChatGPT Routes Follow-Up Calls Across Connectors

ANNOUNCEMENTPRODUCT

A developer reports that ChatGPT silently routed a confirmation follow-up from Connector A through Connector B, carrying a one-time code across separate OAuth clients. The server rejected the mismatched code, but the behavior could enable cross-app data leakage when connector boundaries are not enforced.

// ANALYSIS

This is a serious trust-boundary failure: connector identity must follow the conversation’s authorization context, not merely the shared MCP endpoint.

  • –Two connectors can represent different apps, tenants, or environments even when they use the same MCP server
  • –Silent routing undermines user consent because the confirmed widget and executing connector can differ
  • –Server-side OAuth binding prevented this reproduction from succeeding, but not every connector may enforce that check
  • –Developers should bind confirmation tokens and sensitive state to connector identity, OAuth client, and tenant—not just server URL
// TAGS
chatgpt-apps-sdkmcpsecuritytool-useapi

DISCOVERED

1h ago

2026-11-07

PUBLISHED

3h ago

2026-11-06

RELEVANCE

9/ 10

AUTHOR

Aldipower