69% of Enterprises Expose Security Risks by Sharing API Keys Among AI Agents

• •69% of enterprises share API keys among AI agents, increasing security risks.
• •A compromised agent can grant attackers access to multiple workflows and permissions.
• •Only 21% of organizations have visibility into AI agent operations, complicating breach detection.

A VentureBeat survey found that 69% of enterprises allow AI agents to share credentials like API keys, creating significant security vulnerabilities. This practice means that if one agent is compromised, an attacker could gain access to the combined permissions of all agents using that key, eliminating forensic accountability. The survey, which included 107 organizations, revealed that only 21% of companies have runtime visibility into their AI agent operations. The shared credentials can lead to various attack vectors, including prompt injection and privilege creep, particularly in crypto and DeFi contexts where compromised agents can irreversibly drain funds. Current industry reports indicate that 45.6% of teams still rely on shared API keys, highlighting a security flaw in AI deployments.