What is AI Governance?
AI governance is the set of rules an organisation sets for what its AI systems may decide and do, and the means of showing they stayed within them. Written policy is where it starts. It becomes enforceable when the rules are stated precisely enough for a system to check an action against them before the action happens, and when every answer is recorded.
From policy document to computable authority
Most governance lives in documents: a credit policy, a refund policy, a risk appetite statement. People read them; systems do not. NOMOS turns a document into a sealed artifact: each rule stated as a condition and an outcome, citing the sentence it came from, and signed so that any change is detectable. A system or an agent asks it before acting, and gets AUTHORIZED, DENIED or ESCALATED with the rule that decided.
What written policy misses
Practice drifts from policy. Officers make informal exceptions, thresholds are tuned, edge cases are handled differently from branch to branch. With decision logs, Studio compares what the policy says with what was actually decided, and tests whether a difference changed outcomes, so an organisation can choose to write the practice down or stop it.
Rules that come from law
Some rules come from regulation rather than internal policy. Article 22 of the GDPR, for example, gives people the right not to be subject to a decision based solely on automated processing that significantly affects them, subject to exceptions and safeguards. Where a law sets rules about decisions, NOMOS can hold them in the same form as internal policy, so one check covers both.
Example
A bank's approved credit policy sets a maximum debt-burden ratio (DBR) of 50%.
Loan officers approve some applications above 50% for government employees, by informal custom. Nothing records which rule was applied, so when an auditor asks why those loans were approved, the bank reconstructs it from emails.
The policy is sealed with the 50% rule citing its clause. Each decision is checked against it and recorded. Studio's comparison with the decision logs shows the government-employee exception, and the bank either writes it into the policy, so it is checked and recorded like any other rule, or stops it.
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.