HighAgenticUnsupervised Autonomy●Realized
Oversight Gap in Autonomous Execution
Agent AutonomyDescription
Agent executes significant action sequences without policy checkpoints, human review, or approval gates. Excessive independent execution leads to compounded unchecked errors and unauthorised outcomes.
Example scenario
Financial agent executes a 47-step trade sequence including large position reversals without any human approval gate.
Real-world evidence●Realized
Insufficient human oversight of production AI systems has caused confirmed real-world harm across multiple industries: court-accepted customer commitments made by chatbots without human review, biased automated hiring decisions affecting thousands of candidates, and financial advice dispensed without licensed oversight. These are documented cases, not hypotheticals.
Primary mitigations
- Mandatory human checkpoints at defined action horizons
- maximum autonomous action budgets
- escalation protocols for high-impact operations.
Detection signals
Autonomy Without Oversight Ratio; checkpoint compliance rate.
Mitigating controls
6 Dual coverage
ZYC-HOV-01Agentic
Checkpoint policies for autonomous execution
Human Oversight
ZYC-ACCT-001Non-Agentic
Human-in-the-Loop
Accountability & Oversight
ZYC-ACCT-002Non-Agentic
Escalation Protocols
Accountability & Oversight
ZYC-OREL-001Non-Agentic
Inference with Human Oversight Rate
Over-reliance
ZYC-SEC-011Non-Agentic
Action Oversight
Security
ZYC-SEC-012Non-Agentic
Resource Abuse
Security