Toxic Combinations: When Cross-App Permissions Stack into Risk

Researchers disclosed that Moltbook left a database publicly accessible, exposing 35,000 email addresses and 1.5 million agent API tokens across 770,000 agents; private messages contained plaintext third‑party credentials (including OpenAI API keys) stored alongside tokens that could be used to hijack agents. The article uses this breach to illustrate a class of risk called 'toxic combinations'—runtime bridges between apps created by AI agents, connectors, or OAuth grants that evade single‑application access reviews—and recommends continuous cross‑app inventorying, bridge review, token hygiene, and dynamic SaaS security monitoring to detect and remediate these exposures.