DemoPricing

What is Human Oversight in AI Decision-Making?

Human oversight is the ability of a person to understand, intervene in and, where it matters, decide what an automated system does. It is meaningful when the cases that need a person reach one before the action takes effect.

What the EU AI Act says

High-risk AI systems must be designed so that natural persons can effectively oversee them while they are in use (Article 14(1)), and deployers must assign that oversight to people with the necessary competence, training and authority (Article 26(2)).

How NOMOS routes cases to people

A rule can escalate a decision to a named role, with a deadline. An agent's call can be held until enough of the right people approve it; the approval covers that exact call, once, and is given by a signed-in person, never by an API key. Everything else proceeds without delay.

The right to contest

Article 22 of the GDPR gives people the right not to be subject to a decision based solely on automated processing with significant effects, subject to exceptions; where those exceptions apply, they have the right to obtain human intervention and to contest the decision. A decision that names its rule can be contested on its terms.

Example

Scenario

A lender wants to automate most personal loan decisions.

Without governance

Either every case is reviewed, which defeats the purpose, or none is, and borderline cases go through unseen.

With NOMOS

The sealed rules decide clear cases and escalate the ones the policy reserves for a person (self-employed income above a threshold, for example) to a named role with a deadline. Both kinds are recorded with the rule that routed them.

Frequently asked questions

Related concepts

See it in practice

Upload a policy, confirm its rules, seal them: any system can then check an action against them.

Open Studio →