AI governance
The risk comes from the model: it can leak data, follow an instruction hidden in something it read, or step out of the role you deployed it in. Governance starts with a list of what you run, and that is where most programs stall.
A check has to sit outside the agent
The right check depends on the decision. A payment threshold or a sanctions screen needs a strict rule that gives the same answer every time. Whether a support reply broke your tone policy, or an assistant stayed in its role, needs a model that reads your rulebook.
We run both, and everything in between, hosted or on your own hardware:
- a rule;
- a small trained classifier;
- a check on a criterion you typed this morning;
- a judge model reading your whole policy document.
Most deployments use several, and the cheap ones decide what reaches the expensive ones.
Any of them is a control when four things are true:
- it is declared outside the thing it guards;
- it is versioned;
- it can be tested before it is promoted (put into service);
- what it decided is recorded.
The same judgment written into the agent's own system prompt is a hope, because the agent you are asking to obey the rule is the one that was talked out of it.
The list of what you run comes from the traffic
Ask an organization to list every AI system it runs and you get a spreadsheet somebody updated in March. Ask which of them can move money, read customer records, or send email, and you get silence.
Here an agent is added to the registry by its first call through Swiftward, so the registry stays current. What is on the record for each agent.
We produce the evidence. Your system's risk tier under the law is your decision, made with your counsel: see our read of the EU AI Act.
A control with no record is only a claim
An auditor does not take "we block prompt injection" for an answer. They ask for the decision record: the rule that fired, the frozen version that was live, and the request you can point at.
Where each question is answered
| What am I running? | Registries and documentation |
| Who may call what? | Authentication and authorization |
| What arrived from outside? | Injection detection |
| What must not leave? | Data redaction · code fingerprinting |
| Did it stay in role? | Role and judge checks |
| What is it costing me? | Spend and loop limits |
| What did it actually do? | Audit and evidence |
Where it applies
Every industry that runs AI: financial services, digital assets, insurance, healthcare, legal, platforms carrying user content.