Researchers urge real-time monitoring of AI agents' actions

Researchers studied the four most widely used AI agent frameworks and found major observability gaps, including missing persistent Agent IDs and records of human approvals. Amid incidents involving OpenAI, Anthropic, Google and Meta agents, they call for telemetry standards via OpenTelemetry before agents scale.

• Four most widely used agent frameworks were studied
• Persistent Agent IDs and human approval records are missing
• Proposed records cover what happened and how agents decided
• Observability standards must be set before agents scale