Glossary

Data redaction

Data redaction removes named kinds of data from content before it reaches someone who should not see it. For AI agents it applies to documents on their way in and to answers on their way out of the network.

In a run

The agent drafts a summary for a client. Before it leaves the network, Ares checks it and takes out what your rules name. The step shows as redacted, and the record names the filters that fired, the parties and the time.

Our word for it

Checked before it leaves.

An example

In the document review run, agents read a services contract and summarise it for the client. The summary carries client names, account numbers and addresses. They are redacted on the way out, and the summary still goes. One agent also reaches for an outside page, and that fetch is refused.

Why on the way out

An agent may need client details to do its work. The person who receives the answer may not. Redaction on the way out lets the agent read what the job needs, while the answer carries only what the recipient should see.

You name what is redacted

You choose the filters and what each one removes. Each redaction on the record names the rule that fired, so a reviewer can see why text is missing.

Redaction and where data lives

Redaction controls what leaves in an answer. Where processing runs is a separate choice: see data residency.

Where it stops

Filters cover stated patterns and languages. A kind of data no rule names is not removed. Redaction does not check whether the summary itself is correct.

Related terms

  • Content inspectionContent inspection checks what flows into and out of an AI agent against stated rules.
  • Data residencyData residency is the rule that data must be stored and processed inside a set country or region.

Tell us about one workflow.

We will say whether an agent should do it, and what the first step costs.

A senior member of the team replies within two business days.